specs/023-verified-publication-corrections/evidence-history-through-20261020-4164dd6.md

Rama main · réplica de lectura del repositorio original.

# Evidence history: Verified publication corrections

Historical implementation and verification records retained from `tasks.md`.
The [task list](tasks.md) owns current status, remaining work and delivery requirements.
Earlier CI waivers are historical and have been revoked. No historical source test
or private experiment substitutes for current runtime or live acceptance.

## Evidence

- Baseline PR #223 at `d0120332b5529e2f9d3c751a2e5ffce48b4945f6` fixes pack capture generations and aggregate evidence, not actual publication identity. Both image publications and all PR/main gates pass; its native helper is now deployed and its worker container is recreated without startup, as recorded below.
- Read-only evidence shows five newly present Heavy Metal files repeat already corrected 1977 payloads under a newer-series claim. Full current-library path ownership and exact acquisition provenance remain unverified; the observed snatches are not an exact release-to-file chain. Existing correct files and private historical repair evidence remain retained.
- T001: independent contract and failure reviews closed bootstrap/crash ordering, canonical page-token semantics and worker same-revision owner drift. Native-held-first rollout order and concrete legacy/manual/replay/writer-cycle boundaries are assigned. All 13 functional requirements and four success criteria map to ordered tasks; the 10-item spec-quality checklist passes. The payload foundation is recorded separately in T002; no registration, repeat relocation or full-feature rollout/final acceptance is claimed.

- T002 completed: isolated payload foundation has 17 passing controls without skips, including independently fixed payload vectors, authored RAR4/RAR5/7z payloads and corruption, ZIP directory-type disagreement, complete plain-TAR termination, held compressed wrappers, bounded source hashing and child output/timeouts. Peer findings on directory indicators, growing sources and compressed trailers were corrected. Both independent delta reviews are clear and staged `make ci-local` passes, including all repository checks and history secret scanning. The actual pinned native-image build gate passed all 17 payload controls without skips at source `3abbe7d` (PR #224, build job `111486330407`); the complete isolated Mylar image gate passed. A separate upstream-source selector failure was corrected by preserving the first native `FROM`; all checks on the corrected delivery head passed before PR #224 merged at `f6c6d8db131c3dfd7e8edc1e18d5506389fa6eb3`. GitHub/local/Gitea synchronization and native image publication passed. No registry or publication-path integration is claimed.

- The resumed image-only backup passed independent verification: all 27,314 records match the complete recorded census and both backup/restore copies; 64 isolated database checks passed, preserving baseline IDs for 2,507 comics, 15,341 issues and 141 annuals. The verified baseline remains private. This proves image-rollout recoverability only; correction enforcement, unique current-library ownership and final cleanup acceptance remain open.

- Access resumed on 2026-10-04; verified image rollout: Mylar is healthy on the published `f6c6d8` payload foundation with the reviewed native pack-generation helper; the `d012033` worker image is recreated without startup and remains held. Komga and NZBGet stayed healthy. Fresh post-rollout SQLite checks preserve every baseline ID, all comic/issue/annual path-binding fields and annual deletion flags, prior Downloaded statuses and the five correct 1977 Heavy Metal owners. All 15 requested archive hashes and attributes match before/after. The fresh workflow contains 35 pack records, all exactly unchanged by rollout; the additional record is outside the earlier 34-pack import acceptance and requires separate verification. Correction registration/enforcement, current-library owner validity, potential rename proposals, repeats, final acceptance and cleanup remain open. The operation backup remains retained while these gaps are unresolved.

- T003 source progress: read-only immutable attestation validation binds exact table/issue/parent/release owners, complete bounded census and revision history, rejects contradictory payload owners and checks an owned final marker against the existing database identity without Store construction or file recreation. The versioned full workflow bootstrap projection preserves unknown records and detects protected additions/removals/nested changes while excluding only documented volatile envelopes/observations and registration intents. Isolated controls cover missing/replaced/corrupt authority, prepared/mismatched markers, annual release IDs, malformed SQLite values and complete census beyond recent-history pagination. T003 remains unchecked: authenticated epoch initialization, durable prepared-marker/SQLite/final-marker writes, writer identity binding and crash/recovery/startup admission are still required. The 28 payload/registry controls and seven workflow controls pass locally without skips; independent review findings on SQLite BLOB values/namespace evasion, combined owner bounds and read-only WAL sidecar creation were corrected and covered. The same 35 controls pass without skips in the published native Python 3.10 runtime using read-only candidate source, no network and isolated temporary state; both independent delta reviews are clear. No live registration or prevention is claimed.

- T003 bootstrap source progress: explicit digest-bound epoch plans and finite acceptance/outcome receipts now bind the existing database/writer identities, exact workflow schema, protected workflow census, unchanged pending fences and reviewed backup/restore attestations. Raw writer → workflow lock → SQLite ordering avoids native replay. Durable accepted receipt, prepared marker, atomic census/committed witness and final marker have exact restart/abort controls; aborted tokens stay terminal, and lost/foreign/malformed markers or census records are held. Real process exits cover seven boundaries; an uncommitted SQLite write leaves a journal and is deliberately held without changing its database/journal bytes. The 43 payload/registry/bootstrap controls and seven workflow controls pass locally without skips, including event-writer lock ordering, foreign-state refusal and final-marker loss. The same 50 controls pass in the deployed native Python 3.10 image with read-only candidate source, isolated temporary state and no network. Both independent current-source reviews are clear. T003 remains unchecked: authenticated isolated rollback-journal recovery, registration durability, startup no-replay admission and production API/helper installation remain required; no live epoch, correction registration, media repair, worker resumption or full current-library acceptance is claimed.

- T003 journal recovery source progress: exact private database/journal/marker captures are retained and independently restored through SQLite before an explicit immutable recovery plan can be accepted. Complete restoration equality includes all workflow records, update envelopes and events. Accepted recovery preserves the original database inode and media fences; ordinary bootstrap remains held until the recovery receipt is complete. A changed interrupted pair requires a new reviewed plan linked to the prior accepted plan and the same independently verified contents. Real process exits cover capture, restore, plan, acceptance, original recovery, completion and marker clearance, plus interruption inside the rollback transaction. Hot page-spill and valid cold journals, contradictory journal state, stale/replaced inputs, capture/receipt tampering, bounded history and malformed or plausible contradictory restoration claims have controls. A review finding that malformed evidence could consume the original journal before rejection was fixed by strict nested validation and retained integrity/complete SQL checks before acceptance or original SQLite access. All 58 publication controls, including 15 recovery fixtures, and seven workflow controls pass locally; the 58 publication controls also pass without skips in the deployed native Python 3.10 image with read-only candidate source, private temporary fixtures and no network. Both independent corrected-source reviews are clear. T003 remains unchecked for primary-key authentication, registration durability and startup no-replay exclusion; production helper/API installation and all publication boundaries remain unfinished. No live recovery, registration, media repair or worker resumption is claimed.

- T003 registration witness progress: typed immutable plans avoid the intent/attestation digest cycle by hashing the body without its intent before deriving the attestation and new census. Read-only validation requires complete committed receipt/attestation coverage and an exact predecessor-census chain from the initialized empty epoch. Missing, orphaned, rewritten or foreign committed receipts and accepted unfinished intents hold admission. Separate registration count/byte limits preserve bootstrap bounds and the aggregate namespace cap. Controls include the full 512-attestation history, receipt loss beyond recent-history limits, malformed plans, foreign database/writer/schema bindings and exhausted bounds. All 67 publication controls and seven workflow controls pass locally without skips. The 67 publication controls also pass in the deployed native Python 3.10 image with read-only candidate source, isolated temporary state and no network. Both independent source reviews are clear. The suggested direct wrong-prefix regression is included. T003 remains unchecked for durable registration writes, fresh native owner observation, authentication and startup enforcement. No live registration, recovery, media repair or worker resumption ran.

- T003 registration transaction progress: internal bounded preparation and exact token acceptance now write the prepared marker, atomic attestation/census/committed receipt and final marker while retaining the original initialization witness. The trusted fresh-observation callback runs under raw Writer before workflow LOCK and must exactly match the reviewed inventory and owner facts. It does not provide API authentication or a production native observer. Accepted-old and prepared-old recovery require fresh observations and unchanged protected workflow/fences to commit, or explicit abort. Prepared-new recovery requires the complete committed chain. Missing/foreign/mixed state holds and aborted tokens remain terminal across interrupted aborts. Real process exits cover seven write boundaries. Registration journals remain held with unchanged database/journal bytes because the explicit journal recovery helper is still bootstrap-only. All 77 publication and seven workflow controls pass locally without skips. The 77 publication controls also pass in the deployed native Python 3.10 runtime using read-only candidate source, isolated temporary state and no network. Both independent reviews are clear. Added controls prove that contradictory appends and namespace exhaustion roll back the prospective SQL changes and can explicitly abort to the prior complete census. T003 remains unchecked for registration-journal recovery, native observation, authentication and startup enforcement. No live registration, recovery, rollout, media repair or worker resumption ran.

- T003 registration journal progress: the distinct typed recovery route independently verifies and restores the accepted registration predecessor with its complete prior correction chain and original bootstrap witness. Original database/writer/schema/workflow/fences and the prepared marker must match. Shared capture/restoration mechanics retain exact full SQL equality, original inode preservation and linked review after interrupted rollback. Recovery never completes registration or replays media. A separate registration recovery requires fresh facts or explicit abort. Independent review found that Python dictionary equality could accept boolean/float marker aliases and consume a journal. Canonical JSON comparisons now reject those aliases across registration, bootstrap/registration journal markers and retained receipt predecessors before original access. Controls cover an existing correction chain, genuine hot page-spill, seven process exits, interrupted rollback, route separation, retained tampering and coherently rehashed missing prior authority. All 86 publication and seven workflow controls pass locally without skips. The 86 publication controls also pass in the deployed native Python 3.10 runtime using read-only candidate source, isolated temporary fixtures and no network. Ruff is clear. Both independent corrected-source reviews are clear. T003 remains unchecked for authentication, the production native observer and startup enforcement. No live recovery, registration, rollout, media repair or worker resumption ran.

- T003 connection/status foundation progress: the opt-in existing-only Store mode performs no creation, chmod or schema DDL and binds the original inode/schema on every connection. Missing/replaced cached databases, unsafe files, rollback journals and WAL/SHM sidecars hold before SQLite access. A genuine interrupted page-spill control proves unchanged database/journal bytes. The passive internal status probe acquires existing raw Writer before complete read-only authority checks, revalidates every call, returns fixed sanitized held reasons and preserves all pending media fences without replay. Zero timeout applies only to Writer acquisition. Both independent source reviews are clear. All 93 publication and 16 Store controls pass locally and in the deployed native Python 3.10 runtime using read-only candidate source, isolated temporary fixtures and no network. Ruff is clear. T003 remains unchecked: native startup/scheduler/queue/tick exclusion, authentication and the production native observer still remain. Default callers retain existing behavior until the focused API/installation adapter is complete. No live recovery, registration, rollout, media repair, cleanup or worker resumption ran.

- T003/T004 native observation foundation progress: the internal observer reads complete bounded unfiltered native comics/issues/annuals under caller-owned raw Writer and binds exact current regular/annual parent/release owners to actual archive SHA/signature/inventory. Strict catalog v1 preserves raw path/status/deletion facts and rejects boolean/float aliases. Missing, duplicate or shadowed identities and deleted/inactive path claims hold. Independent failure review found that lexical paths missed symlink aliases and duplicate parents could expand quadratically; corrected metadata identity checks reject aliases, count physical inode owners, recheck every claim and require one parent before expansion. Private uncatalogued backup hardlinks remain valid. Row/byte/path-depth/identity limits and a shared cooperative deadline bound the work. Only selected payloads are read. All 113 publication and 16 Store controls pass locally, and all 113 publication controls pass in the deployed native Python 3.10 runtime with read-only candidate source, isolated temporary state and no network. Both independent corrected-source reviews and Ruff are clear. The production callback fixture proves stale native status rejection before commit and successful fresh reobservation after restoration. T003/T004 remain unchecked for authenticated production wiring, startup exclusion and the full API protocol; T005 installation and every publication boundary remain open. No live recovery, registration, rollout, repair, cleanup or worker resumption ran.

- T004 authenticated protocol completion: `publication_api.py` validates exact version/action schemas, bounded JSON, owner/evidence/census facts and rejects caller paths before constructing state. The checked primary-key-only POST adapter denies SSE, disabled APIs, GET, JSONP and unknown arguments before controller import. Preparation derives current native observations and binds reviewed payload/census; exact token acceptance/recovery uses the SDK without media replay. Passive exact receipt status uses existing immutable read-only SQLite under raw Writer before workflow LOCK. Outcomes use durable receipts, including accepted abort and replay. Independent reviews caught and corrected an abort response label, a backup bound mismatch and a recovery review read after marker clearance. All 17 focused controls pass without skips against pinned source, including actual patch idempotency, corrupted read-only status, stale catalog/census, wrong typed journal route, genuine hot-journal old-state restoration and preservation of its existing prior correction. Both corrected-source reviews are clear. T003 startup and all publication boundaries remain open; no live registration, recovery, rollout, cleanup or resumption ran.

- T005 adapter/image verification completion: the focused adapter installs after writer/workflow/naming dependencies and copies only the owning runtime helpers. The actual deployed Python 3.10 image passed the complete isolated reliability gate with read-only candidate source, no network/live mounts and disposable state, including all 113 publication, 16 Store and 17 API controls without skips plus actual patched package/API imports. Checked patching is byte-idempotent and rejects drift. Both corrected-source reviews are clear. Image build/delivery checks remain part of T016; no live operational action ran.

- T003 startup candidate progress: early publication mode excludes legacy replay before native database maintenance. Existing-only workflow factories, admission under raw Writer, held health/UI/API/schedules/ticks, and restart-before-completion have 39 passing actual-source host controls. Independent reviews found and corrected a wrong DDL JSON Store substitution, swallowed dbcheck failures, notification polling writes, native sidecar appearance during fresh preparation, missing native catalog recreation and zero-byte catalog admission. Fresh creation requires explicit reviewed evidence, an exclusive retained claim and exact accepted bootstrap receipt; its one-time native creation permission is consumed before dbcheck. Existing catalogs require a bounded complete readable rollback SQLite header and native base tables with unchanged signatures. The frozen-source full native gate v5 passed all 113 publication, 16 Store, 17 API and 39 startup controls without skips. Both corrected-source reviews are clear; Ruff, repository validation and full local CI pass. T003 source acceptance is complete; matching PR/main image delivery is the next gate. All payload boundaries, live rollout, reconciliation, cleanup and worker resumption remain open. No live mutation ran.

- T003 delivery: PR #235 source `268a23a92a40b53676b698639646c0b2cf9a7fe8` merged as `a8bfb7e900b4e2a17e5b035b83d311e417167e2a` after all 13 checks passed and no unresolved review threads. PR and main actual-image builds passed 113 publication, 16 Store, 17 API and 39 startup controls without skips. The full-revision GHCR OCI/config and source label independently match main build digest `sha256:f96d2ad2d7645ff6ceb2a2ee3d0a3368debdc1a90146eea33f043e26fe2c5fa9`. Local main and the configured Gitea mirror are synchronized. Main lint, validation, CodeQL and image publication passed; Pages remains pending behind a queued earlier deployment. No live startup rollout or prevention acceptance is claimed.

- T006 source acceptance: the checked adapter installs 47 guards across actual ordinary/annual/storyarc/oneoff scripts, tagging, duplicate handling, placement, complete tidyup deletion scope and catalog/status/history boundaries. Fresh local payload/full-census/current-owner checks preserve unfiltered annual parent/release identity; synthetic IDs confer no registered owner. Retained review bypasses legacy failure fallbacks and is acknowledged without false completion or failed-download replay. Call-specific ARC/oneoff/multiple operations, same-path softlinks, registered original relocation/deletion, unclassified cache files, changed cleanup entry sets and initially absent originals/cache paths are guarded before mutation. All 36 host controls and the frozen actual Python 3.10 native image gate passed without skips, alongside 113 publication, 16 Store, 17 API and 39 startup controls, actual patched package imports and the complete existing reliability gate. Both corrected-source reviews are clear; focused Ruff, `make validate`, staged `make ci-local` and secret scanning pass. Registered original moves/deletions remain held until the later verified transition contract. Source delivery is the next gate; T007–T019 and matching live prevention/recovery remain incomplete. No live mutation ran.

- T006 delivery: PR #236 source `5be86cf473ce25b5c15ba1f21a5fe10fd7da54ed` merged as `47e9d284c1f413cff1d87a9bdda902a5c3cc5171` after all 13 checks passed with no unresolved review threads. PR/main actual image builds passed 113 publication, 16 Store, 17 API, 39 startup and 36 postprocessing controls without skips. Independent full-revision GHCR OCI/config/source verification matches main build digest `sha256:a87884491dbcd1bd9921c820e37f823b393f9066698d00d22538cc1a222e45d3`. GitHub/local/Gitea main are synchronized, all five main workflows including Pages succeeded and no open PRs remain. A superseded queued Pages run was cancelled to unblock current documentation delivery. No live rollout or prevention acceptance is claimed.

- T007 candidate progress: added fresh Writer-held payload/owner checks before either backend and after returned temporary paths, distinct manual retained-review responses and lower-boundary review recording before the processing observer unwinds. Initial backend controls exposed the missing guards; the corrected entry controls now cover rejected/synthetic owners, unavailable registry, changed payload and correct/distinct eligibility. This is only producer-entry progress: exact owned-job coordination, direct Modern/Service/converted paths, mutation-capable receipt replay, final publication/cleanup and successful real backend controls remain required. A generic nested-fence bypass is prohibited; T007 stays unchecked.

- T007 owned-job prototype: a private typed intent binds raw Writer identity, token, actual source signature/SHA/payload, exact owner, captured policy, complete census, current matched-owner facts and exact pending fence. Modern capabilities additionally bind all four existing private recovery directories and their private identity file; replacing either holds the job, while expected child creation remains possible. Only the active in-process capability can request internal payload checks; ordinary admission/advisory calls remain held and interrupted intents cannot initialize another job. The first full native regression passed 20 tagging controls alongside 113 publication, 16 Store, 17 API, 39 startup and 36 postprocessing controls without skips. This supersedes the earlier host API invocation's source-dependent skip. The prototype performs no receipt replay or fence clearance. Producer integration, final mutation/receipt/cleanup hooks, safe exact-fence completion and real successful Legacy/Modern/converted controls remain unfinished. No live action ran.

- T007 direct/converted candidate progress: direct Modern entry verifies the actual payload before catalog lookup, state preparation and fence writes; manual conflicts return retained review. Converted follow-ups compare the original conversion SHA before acknowledging existing ComicInfo. Production polling owns the native Writer before Store construction and retains it through selection, processing and durable review writes. Current payload/owner checks precede receipt/metadata shortcuts and repeat immediately before completion; terminal review retains attempts/token without retry. Missing-owner catalog reconciliation cannot reassign a registered payload from its filename, and unbound receipt capture is held before recovery-state preparation. A changed prepared intent is retained instead of overwritten during fencing. Independent review caught and corrected the production polling lock seam and relative Legacy output paths; successful relative outputs now become independently verified absolute paths before caller copying. All 31 host tagging controls and 14 converted queue controls pass. The frozen earlier image gate passed 29 tagging controls and the full reliability suite without skips; the corrected-source review found no actionable issues and independently reran all 31 tagging and 14 converted controls without skips. Final native verification for the two relative-output controls remains pending. These are guards, not successful owned publication/replay acceptance; T007 stays unchecked.

- T007 Service and lower automatic-publisher candidate: exact active Modern jobs bind the actual Publisher/cache/staging/receipt/configuration roots, captured policy/token and complete source/correction facts. Unbound receipt replay and late appearing token receipts remain retained. Publisher receipts carry the exact owned intent/source/payload/census binding; checks precede pack updates, NFS rename/link/unlink/restore and complete workspace cleanup. Automatic staging alone may temporarily lose its name while its exact displaced original remains verified; the acquisition and registered library copies remain intact. Sixteen genuine Service controls pass without skips, including actual v2 no-overwrite and metadata-output jobs, owner drift during the CLI and after displacement, unknown workspace retention and registered in-place refusal before library displacement. Independent review is clear for these implemented seams. Holding the original pending-marker descriptor now prevents inode recycling during the job; all 15 transaction controls pass. The broader host gate passed 259 tests with two explicit image-dependent skips (actual native API adapter source and the pinned ComicTagger executable); `make validate` passes. Current-source native verification, controlled changed in-place transitions, explicit interrupted-job replay, producer integration and exact terminal fence completion remain unfinished. No live prototype rollout or prevention completion is claimed; T007 stays unchecked.

- T007 native producer and completion candidate: the actual Modern entry now prepares an exact owned transaction before its pending fence, supplies its private coordination capability to Service and verifies terminal Publisher/Staging evidence before removing only its captured marker and intent. Automatic temporary output and manual no-overwrite handoffs preserve the acquisition and page payload. Interrupted marker/intent removal retains a typed hold; failed final intent-directory fsync restores that hold exclusively without overwriting foreign evidence. Backend output checks also bind unknown input/output payload lineage, rather than merely accepting two independently unknown archives. Review found that final native admission could replace a retained-review result with a generic tagging failure; the correction propagates typed review through every owning operation while still running admission and refusing the next held operation. Processing acknowledgement occurs under Writer before review crosses the admission context. Four current genuine owning-admission controls pass without skips; the earlier twelve backend payload controls pass. Current frozen native reliability and expanded host checks are running. Changed in-place publication, explicit interrupted-job recovery, converted successful publication and complete matching-image/live acceptance remain unfinished; T007 stays unchecked. No live prototype rollout or library mutation ran.

- T007 closed-history candidate: an exclusive private terminal witness permits reading an exact previously completed receipt without replaying it or requiring its disposable output to remain. A second genuine native job passes while preserving the first receipt; altered or missing witness/receipt evidence remains held. Review identified and corrected missing full-job and terminal-fact bindings: the retained fenced predecessor uses a canonical intent digest, historical correction epochs/revisions are checked, and the current Publisher receipt anchors the complete terminal projection before owned fence clearance. Witness evidence is checked through both release boundaries. Disabled automatic tagging and unsupported Modern source types preserve their existing eligibility before a job is prepared; recovery-state errors become terminal review under Writer. Expanded history/fidelity/policy checks and corrected-source review remain underway. The earlier frozen image passed 66 tagging controls and the existing publication/native suites, but its final pinned-SDK installed-module comparison correctly rejected mounting new modules on the old image; a separate actual patched candidate image is required. `make validate` and full local CI passed before this closed-history extension. Source, image, changed in-place/replay and live acceptance remain open; T007 stays unchecked.

- T007 final review correction: independent failure review reproduced Legacy fallback cleanup creating a `.BAD` file beside a missing original before a postcheck could refuse the job. Publication mode now returns retained review before Legacy copy, child or cleanup; existing behavior outside publication mode is preserved. All 13 current backend controls pass without skips, including automatic/manual refusal with unchanged archives, database and fences. Both corrected-source reviews are clear. The prior actual patched image passed the complete reliability gate, all 79 tagging controls and all 18 real pinned ComicTagger SDK controls after its stripped service fixture loaded the actual Writer module. The new frozen candidate includes the Legacy refusal and 80 tagging controls; its complete actual patched image gate passes all 80 tagging controls, all 18 real pinned SDK controls and the existing reliability suite without skips. Current staged `make ci-local` passes with no optional-tool skips. Legacy owned tagging, changed in-place publication, explicit interrupted recovery, converted successful publication and matching live prevention remain unfinished; T007 stays unchecked. No live prototype rollout or library mutation ran.

- T007 CI fixture correction: the upstream source-only image does not install the archive verifier used by the new publication fixtures. These 80 controls now explicitly skip in that source-only gate; its actual pinned base gate passes. The custom image installs the verifier and its existing required-tool gate remains enforced; all 80 tagging controls and 18 real SDK controls passed without skips on the frozen candidate. Full staged local CI passes. T007 and live prevention remain open.

- T007 retained-copy progress: the owned intent and Publisher receipt now bind the complete before/after NFS transition before replacing a retained copy outside registered correction owner paths. Registered original updates remain held. An actual converted catalog/queue/native/Service chain completes a distinct eligible publication once while preserving the prior registered archive. Four focused controls pass without skips for successful retained-copy/converted publication, owning-admission review and expired-capability refusal. Separate self-review reproduced passive proof recreating a lost journal; it now uses only its already-bound publisher and refuses closed capabilities before access. The broader 83-control host run found one obsolete review fixture targeting a newly eligible retained copy; that control now targets the registered archive it is intended to protect and passes. The frozen actual patched image passed all 85 tagging controls, all 18 pinned SDK controls and the complete existing reliability gate without skips, including real CLI retained-copy publication. The host passed all 85 controls with one explicit pinned-executable skip; the same control passed in the image. Legacy owned tagging, registered source transitions, explicit interrupted recovery and remaining writer boundaries stay open; T007 remains unchecked. No prototype live deployment ran.

- T007 interrupted-binding correction: a one-shot receipt write failure reproduced ordinary tagger recovery removing the prepared workspace while the partial correction intent remained held. Binding durability errors now cross the caller as typed retained review before fallback recovery or cleanup. Three fault controls cover intent, receipt and activation writes and require both original archives, prepared recovery copies and the admission hold to survive. The four focused fault/success controls pass, and the frozen actual patched image passed all 88 tagging controls, all 18 pinned SDK controls and the complete existing reliability gate without skips. Full local CI passes with no optional-tool skips. PRs #238 and #239 are merged under the user's explicit GitHub CI exception during its hosted-runner incident; the complete original branch protection was restored after each merge. GitHub/local/Gitea main match at that delivery point. Current main integration preserves the calendar fix and both sets of task evidence. No prototype live deployment ran. T007 remains unchecked.

- T007 publication receipt durability progress: a new fault control reproduced fallback recovery after an owned publication receipt write failed, outside the previously guarded in-place binding writes. Owned Publisher writes now return typed retained review before fallback recovery or cleanup. Initial receipt, publishing receipt and terminal commit failure controls retain the appropriate prepared/displaced originals and admission hold, including post-replacement errors. The prior frozen image passed all 93 tagging controls, all 18 pinned SDK controls and its complete reliability gate without skips; its host gate passed 93 controls with one explicit executable skip. An additional automatic-path control passed on the same installed runtime with exact module hash equality. Full local CI passed before the registered-source extension below. These results precede the larger combined T007 batch and do not accept its later source. No live prototype deployment or library mutation ran.

- T007 registered-source batch progress: the private active capability can now resolve its exact catalog name to its verified before/displaced/after NFS archive. Full current catalog claims and physical identities remain checked, and only the exact bound metadata transition may project its source hash/signature onto the captured observations. A real native registered metadata job preserves payload and catalog ownership and completes. Direct owned Service completion, owner drift after displacement and an intervening deleted annual shadow pass; both drift cases retain displaced and original copies with admission held. All 21 observation controls pass, including refusal of boolean/dictionary/foreign capabilities before catalog mutation. Expanded host, pinned-SDK and matching-image verification for this combined batch remain pending. Legacy owned tagging and explicit interrupted recovery remain open; T007 stays unchecked. The user requested larger implementation batches, so this work is being grouped before delivery rather than opening a receipt-only PR.

- T007 combined batch source progress: registered metadata transitions and their released immutable handoffs now preserve fresh complete owner/catalog facts; owned Legacy CBZ/ComicRack tagging uses the pinned offline 1.3.5 parser with private configuration and refuses lossy policy/metadata. Explicit terminal tagging completion now binds a reviewed backup/restore attestation, fresh archive/receipt/staging/owner/full-census snapshot and immutable witness to a private completion token. A separate hold and durable accepted ledger span fence and intent clearance; completion never tags or replays media. Earlier or unwitnessed phases remain held. Automatic recovery and focused image Legacy/registered controls have passed; the expanded interruption controls and matching final image/full local gate remain pending. T007 stays unchecked until the combined batch is accepted. No live prototype deployment or library mutation ran.

- T007 combined source acceptance: all 115 tagging controls are verified against identical frozen production modules in isolated images: the broad V8 run passed 113 and exposed one superseded registered-source hold expectation plus one fixture missing its active runtime owner; the corrected two controls pass in V9. Only those two test fixtures differ between candidates. All four Legacy controls (including the actual bundled parser), all 18 pinned Modern SDK controls, ten failed-download controls and the remaining image streaming checks pass without skips. Installed changed production modules equal the frozen source. The new explicit recovery controls cover automatic/manual completion, each hold/acceptance/fence/intent/completed interruption, owner drift, changed output, missing witness, unwitnessed phases, marker-only native admission, final-sync hold restoration and redacted API responses. Repository validation, focused Ruff and staged local CI pass; the host broad run is supplemental and retains explicit SDK skips. Staged self-review found no remaining actionable defect; this is not independent review. This completes the T007 source boundary only. T008–T019 and spec 022 current-library acceptance remain required; no prototype deployment, live media write, registry seeding or normalizer activation ran. GitHub delivery will use the user's persistent CI waiver while preserving/restoring the original protection.

- T007 delivery: PR #241 merged the combined owned-tagging batch at 41a0d76aae32e2a3ff048046dd748e1df04335ca under the persistent user CI waiver. The full original branch protection was independently compared and restored exactly. Local/GitHub main and the actual Gitea main match the reviewed tree. The supplemental host run executed 115 controls with the same two corrected fixture failures and three explicit SDK skips; the two corrected host controls pass. Image controls remain accepted across byte-identical production modules as recorded above. No dependency contract changed and no prototype live deployment ran.

- T008/T009 grouped boundary progress: rescan parsing now accepts a trusted native publication callback before exposing parser corrections, and the checked actual updater patch installs it before duplicate/catalog changes. The callback checks existing protected catalog-bound archives even for an empty scan, so a missing correct archive cannot silently archive or clear its rows; wrong known payloads and deleted annual shadows retain review while genuinely different archives remain eligible. Converted-catalog updates check fresh publication ownership before their conditional row update. Focused real-registry/archive and owning parser/catalog controls are running. Actual-image acceptance, remaining manualRename/downloadLocal/movefiles/file_ops/libraryScan mutation branches, naming/supplement/replay/derivative contracts and T008/T009 completion remain pending. Registered-original movement remains held until verified relocation can update its catalog binding safely.

- T008/T009 actual rescan acceptance: the frozen rescan V2 image passed 92 rescan/catalog/request controls and 14 actual native file-matching controls without skips. The installed helpers equal frozen source, and checked patch migration from the prior rescan marker is idempotent. The grouped transfer extension adds fresh payload, exact proposed owner, current unfiltered participant-path and physical-alias checks before actual manualRename/downloadLocal/movefiles/file_ops mutations and libraryScan archival row updates. Ordinary imports with genuinely different payloads remain eligible; protected catalog-bound relocation and overwrite are retained review. First transfer-image verification exposed two fixture errors (the destination-directory test used a different output basename, and the native hardlink assertion used incorrect source spacing); corrected fixtures require another frozen-image run. T008/T009 remain unchecked pending acceptance and the remaining naming, supplementation, replay and derivative contracts. No live media mutation, registry change or worker activation ran.

- T008 source acceptance: the frozen transfer V11 image passes 153 focused controls without skips, including the actual native file-operation function and patch idempotence on an isolated native tree. Rescans hold changed correct-owner statuses and missing originals even when enumeration is empty; transfer checks preserve both current participant bindings and attested correct paths after a catalog binding is lost, reject directory-wide transfer inference, physical-alias overwrites and rollback-journal/WAL uncertainty, and admit correct copies, unchanged-path moves and genuinely different unbound archives. Native outer requests return retained review without success acknowledgement or failure search. Rescan V2 additionally accepted all 14 actual native file-matching controls. The full transfer V9 reliability run remains in progress; its production delta to V11 is the catalog recovery-state hardening covered by the current focused run. Earlier V6/V8 failures were isolated fixture imports, missing active processor context and attempting to patch a read-only native tree; the corrected fixtures now pass. Repository validation, focused Ruff and staged local CI pass for the pre-weekly-fix batch. T009–T019, independent review and live publication acceptance remain pending. No publication prototype, correction registry or normalizer activation ran.

- Grouped T008 acceptance closure: the frozen transfer V9 full reliability gate passes, including all 115 tagging controls, four bundled Legacy controls, 18 pinned Modern SDK controls, ten failed-download controls, streaming discovery and actual package imports. Current V11 acceptance adds 153 focused controls without skips for catalog recovery-state hardening and the separately requested weekly-date repair. A weekly feed date conflict still returns `incorrect_match` but cannot turn an existing Wanted issue into Skipped; the UI describes release review while preserving stored status compatibility. Its actual updater/template tests pass, and the control rejects the original status mutation. This completes T008 source verification, not T009 or live publication prevention. The separate live weekly hotfix is built from the prior live image with only `updater.py` changed among 136 package modules; its backup/restore and scoped operational acceptance remain pending.

- Live weekly-date acceptance: a disk-backed, quiesced application-state backup and independent isolated restore verified 20,735 files and 32 SQLite integrity checks before rollout. The scoped hotfix changes only the native updater and weekly template. Authenticated `/pullist` returns HTTP 200 and describes the date conflict as release review. Absolute Batman #24 retained its correct series/release identity and Wanted status through rollout, then completed a normal Snatched/Post-Processed import. Its current Downloaded archive has 28 readable pages, matching issue/series metadata and an exact release-ID link. Only its status, location and size changed after the accepted rollout baseline. All 136 deployed native Python modules match the frozen hotfix. The original 2,516 series, 15,443 regular issues, 142 annuals and 3,247 Downloaded/Archived bindings remain preserved. This closes the separate weekly hotfix acceptance. It does not activate the publication prototype or complete T009–T019.

- Retained intake recovery: four uniquely matched Snatched releases were submitted once through the existing source-preserving recovery path after independent backup/restore verification. MAD #39 and the Gambit, Hulk and Spider-Gwen Galactus one-shots now have Downloaded records and library files. Archive verification preserves all 61, 24, 25 and 23 page bytes in order, along with every original non-metadata member. Original intake files and prior worker receipts remain preserved. A fifth Detective Comics collection has no unique catalog match and remains untouched for review. The normalizer remains held. These scoped recoveries do not establish current whole-library acceptance or automatic prevention for unfinished publication paths.

- T009 boundary safety progress: direct SDK tagging/recovery and NFS no-clobber restore now require the existing exact owned native producer before receipt replay or file changes. Fresh naming, legacy naming acknowledgement, reader supplementation and nested metadata repair remain explicit review holds before unbound publication intents or fences. Nested repair records review without spending an attempt or token, and its poll consumes retained review without killing the worker thread. Standalone supplementation checks retained publication history even when its marker is missing. Current frozen-source owning controls pass 130 tests without skips in the actual native image runtime. The host run passes the same 130 controls with four accurately reported executable/verifier skips. Focused source lint, repository validation and all-file local hooks pass. The unchanged-production SDK run also passes all 115 owned-tagging controls without skips, for 245 current owning/SDK controls in the actual runtime. T009 stays unchecked: safe correction-aware naming transitions, reviewed member-name derivative aliases and complete maintenance recovery acceptance are still required. This source batch has not changed the live image, correction registry or normalizer hold.

- T010 read-side foundation progress: the independent worker context now has a generated, explicitly selected native evidence reader with no registry transaction/recovery classes. Repository validation rejects stale copies. The trusted local adapter proves matching config/writer physical identities, projects explicit nonoverlapping media mappings, and checks full census/witness history, current candidate, all-row matched owners and correct archive bytes under caller-owned shared Writer without remote calls. It rechecks authority, catalog and source signatures/hashes and shares one artifact deadline. Twenty-four portable/current-owner controls pass on the host without skips; the full host worker suite passes 234 tests with 72 converter-dependent skips. Focused lint and repository validation pass. The frozen worker image build passes all 234 tests without skips, and all 18 installed production modules match the frozen source. This foundation is not yet called by writer_cycle or mutation paths: T010 remains unchecked, as do T009 positive transitions/aliases and T011–T019. No live media, correction authority or normalizer activation changed.

- T010 cycle boundary progress: coordinated cycles now validate complete local authority before recovery identity/fence writes and recheck its namespace before fence clearance. Explicit publication_roots mappings are mandatory under coordination. The thread-owned adapter cannot be used after writer release. Native Mylar HTTP is refused before credentials/network while the writer is held; a zero-time physical flock probe covers a separate local registry and releases before HTTP. This closes the self-wait hazard without bypassing any pending fence or registry checks. Eight new cycle/HTTP controls supplement the 24 portable/current-owner controls. The full host worker suite passes 242 tests with 72 converter-dependent skips. The frozen actual worker image passes all 242 tests without skips, and all 18 installed production modules match frozen source. Focused lint, repository validation and the final all-file hooks pass. Qualified self-review found no remaining actionable defect in this bounded cycle/refusal change; independent T016 acceptance remains open. T010 remains unchecked for complete accepted cycle/handoff integration. Per-source mutation checks and durable outside-lock API handoff in T011–T012, T009 safe transitions/aliases and all live acceptance remain unfinished. Stack contracts document the required mappings, initialized authority and retained worker hold; no live changes ran.

- T011 grouped import boundary progress: common import now checks fresh full authority, exact unfiltered native issue/parent/annual-release owner and candidate before stage/receipt creation, both original and copy after staging, and again before an attempt receipt. Guided claims and pack candidate/existing-target confirmations check before acknowledgement or replacement facts. Pack cleanup rechecks all confirmed comic sources/destinations before any cleanup intent, report or deletion; old supplement cleanup retains review pending derivative ownership. Sixteen new owning/import controls cover wrong owners, deleted/locationless shadows, missing/coerced parents, owner/authority drift, catalog journals, prior-proof preservation and staged-copy cleanup. The full host worker suite passes 258 tests with 72 converter-runtime skips. The frozen final V3 worker image passes all 258 tests without skips, and all 18 installed production modules match frozen source. Focused lint, repository validation, final all-file hooks and staged commit hooks pass. Qualified self-review corrected missing-parent validation and unregistered target ownership before final acceptance; independent T016 review remains open. T011 stays unchecked for durable API handoff/receipt bindings, standalone duplicate and remaining report/Extras paths; T009 positive transitions/aliases and T010 integration remain incomplete. No live mutation or worker activation ran.

- Common import handoff continuation: the source now gathers bounded immutable native work before the cycle lock, writes private prepared source/stage/owner/census receipts under the writer, and attempts dispatch only after verified fence clearance. Dispatch persists uncertainty before outside-lock HTTP. Native submission checks its actual stage and stores a durable attempt; the queue carries that proof into fresh processing admission with a separate at-most-once processing record. Later publication checks constrain payload and owner. Dispatch requires an exact supported native handoff version before spending a prepared attempt. Native attempts also refuse a different token for the same exact owner until review. The host worker suite runs 267 tests with 72 converter-runtime skips; the host workflow suite runs 44 tests with one native-source skip; host native checks pass nine controls with 34 runtime skips. Repository validation, final `make ci-local`, focused lint and all-file hooks pass for the source changes. The isolated native build awaits pkexec authentication. Matching image gates, source parity, final review/delivery and live acceptance are pending. Guided acknowledgement and pack report/catalog handoffs remain unfinished. No task checkbox or live worker hold was cleared.

- Common import handoff image acceptance: frozen worker source passes all 267 tests without skips, and its 18 installed production modules match source. The native candidate passes the complete isolated reliability gate, including 43 native publication controls and all 115 tagging controls without skips. The four installed native handoff/evidence modules match frozen source. A separate qualified self-review traced authenticated submission, instance queue provenance, processing admission, changed-stage refusal, durable attempt history, fence clearance and preserved source receipts. No actionable defect remains in this bounded common-import change; independent feature acceptance remains open. Source delivery is pending. T010/T011 remain unchecked for their remaining guided/pack/report/Extras and typed reconciliation requirements, and all live holds remain in place. No live deployment or media mutation ran.

- Common import handoff delivery: PR #251 merged at `33c61bf6bbc83888723a4090e84ba36b56db094f` after the complete isolated native gate (1,023 test executions across 70 suites) and 267 worker tests without skips. Four installed native handoff modules and all 18 worker production modules match frozen source. Reviewed and merged trees match, original branch protection is restored, and local/GitHub/Gitea main synchronization is verified. The user waived GitHub CI; local and actual image gates passed. The worker remains held, with guided acknowledgement, pack catalog/report dependencies, safe naming/metadata transitions and final library acceptance still unfinished. No live deployment or media mutation occurred in this source batch.


- Larger guided/pack boundary batch in progress: source now binds guided command/source/version to native queue admission, reconciles a prepared receipt after command-record loss, and dispatches typed pack catalog/report and guided acknowledgement requests outside the shared writer. Pack reports bind private extracted originals locally and validate shared catalog targets natively; stale reports cannot erase verified members, and cleanup requires a durable report acknowledgement. Extras placement holds registered payloads and protected physical originals before metadata changes and uses exclusive temporary creation. Fresh registered-owner confirmation holds changed archive payloads without reviewed lineage. Focused restart, lost-response, old-protocol, source/owner/census drift and protected-original controls pass on the host. Full frozen native/worker image acceptance, final source review, repository gates and delivery remain pending; host converter/verifier skips do not count as runtime acceptance. T009–T012 and live acceptance remain unchecked. No live media, registry or worker activation changed.

- Larger batch verification progress: the full host worker suite passes 280 tests with 72 converter-runtime skips. Ten focused maintenance handoff controls pass without skips, including a further parent-traversal refusal check. Native publication controls using the local pinned verifier pass 50 tests with 14 actual-native-package skips. Native workflow, pack records and catalog suites pass 48, 19 and 10 tests respectively, with one workflow native-source skip. Patch migration is checked for both prior API markers. Review additionally closed native catalog drift before attempt recording, verified-member kind/owner downgrade, dispatch packet byte substitution and differing native/worker catalog mappings. The frozen worker V4 image passes all 281 tests without skips, and all 19 installed production modules match its frozen source. Native V3 exposed an ambiguous fresh-patch locator after the typed acknowledgement insertion. Its corrected command-list locator and monotonic completed inventory pass their owning controls; the frozen native V4 build has passed all 50 publication controls without skips and is finishing the full reliability gate. Earlier frozen contexts are superseded and are not acceptance evidence. No live deployment or media mutation ran.

- Larger guided/pack batch image acceptance: the final frozen native V4 image passes 1,037 test executions across all 70 reliability suites without skips, including the fresh API-patch migration, 50 publication controls, all 115 tagging controls and the bundled Legacy/Modern runtimes. The final worker V4 image passes all 281 tests without skips, and its 19 installed production modules match frozen source. Full local CI and source lint pass. Qualified self-review closed restart reconciliation, real source-root/private extraction mapping, native catalog drift, immutable dispatch bytes, delayed terminal identity/inventory downgrade, malformed history and traversal checks. All eight installed native handoff/authority modules also match frozen source. Source delivery remains pending. Independent feature acceptance, T009–T012 remaining integrations and live rollout remain open. No live media, registry or worker activation ran.

- Next larger implementation batch: extend T011 duplicate cleanup and interrupted receipt recovery together, binding current source/target payloads, native catalog ownership, complete correction census and immutable receipt predecessors before retention, unlink or terminal acknowledgement. Add source/target/census drift, protected-original, prior-proof replacement and interrupted cleanup controls. T011/T012 remain unchecked until their complete boundary requirements pass; this batch does not authorize worker activation.

- Guided/pack source delivery: PR #252 merged as `2426a42b76e261bcc98cc841ecf3499b7e377e5e`; reviewed and merged trees match and original branch protection is restored. The accepted frozen native gate has 1,037 executions across 70 suites, with eight installed native modules matching source; the worker has 281 tests and 19 installed modules without skips or source differences. GitHub CI was waived by the user. No live rollout, registration, media mutation or worker activation occurred.

- Parallel larger batch: separate agents own conversion, naming and reader notification boundaries, with one parent integration/Git writer. Duplicate cleanup now binds exact current native target ownership, complete census and all catalog path/physical-original absence, retains an exclusive original, and uses immutable private CAS receipts. Source/owner/census drift and lost unlink acknowledgements are held. Independent review found and closed an unregistered catalog-original hardlink gap. Naming preparation now binds its reader/publication/preservation facts; conversion paths without typed relocation or reviewed PDF lineage stay retained. Reader notifications use durable outside-writer handoffs and fresh owner/source/census checks. Full combined host/image/repository acceptance remains pending. T009–T012 stay unchecked while their remaining positive publication/recovery contracts are implemented.

2026-10-06 parallel implementation checkpoint: separate file owners implemented
worker duplicate protection and retained unlink acknowledgement, immutable pre-lock
reader observations, reader notification handoff, naming/conversion boundaries,
and typed maintenance diagnostics; native owned rename and preserved root
ComicInfo publication have positive controls. Full worker host suite passed
376 tests with 72 converter/runtime skips. Container gates caught standalone
private-source leakage and an ambiguous workflow patch anchor; fixes and new
frozen image gates are in progress. Neither failed candidate is accepted. The
combined metadata/rename runtime route, derivative transitions, final complete
feature review and live acceptance remain open; task checkboxes are unchanged.

2026-10-06 verified source milestone: the final isolated worker image passed
377 tests without skips; all 21 installed production modules match its frozen
source. Repository validation and full local CI passed. Source commit `8f23324`
and history synchronization `5c90148` preserve the same reviewed tree; PR #253
remains a draft until the final native image and installed parity pass. Native
focused controls passed: 20 owned rename, 28 metadata maintenance, 33 naming/API
and 12 enrichment controls. Independent terminal-witness and unchanged-source
race findings are closed. The legacy tagging host regression passed 115 tests
with three binary-only skips; the actual image must run them without skips.

The next implementation runs in separate worktrees with explicit ownership:
combined rename/unchanged-hash reader restoration/root metadata orchestration;
native-owned lossless container conversion and worker acknowledgement; private
read-only correction-plan tooling. Shared authority and manifests have one
integration owner. No live writer is released and no full task is checked off
from these partial source milestones.

2026-10-06 final image acceptance for PR #253: the isolated native image passed
1,082 test executions across 72 suites without skips, including its actual
Legacy and Modern binaries. All 19 checked installed native modules match the
unchanged frozen source. The worker passed 377 tests without skips with exact
21-module parity. Earlier failed or superseded candidates are not accepted.
The reviewed source milestone is ready for delivery with the user's hosted-CI
waiver; complete feature, combined-route, derivative and live gates remain open.


- Coordinated continuation: PR #253 merged at `6cac5816676500c81edb8f559ab433e50a5a93ba`; GitHub, Gitea and the clean local main checkout match, and branch protection is restored. Separate worktrees now own the combined rename-before-metadata route, native-owned lossless conversion, and private correction plans/reviewed derivative lineage. The parent owns shared authority/API/installation, task evidence and Git integration. The combined pass retains its native original/restore pair; conversion publication preserves member/page inventories and requires exact current native ownership. Independent controls verified changed-source refusal immediately before retirement and final retained-original/witness refusal before acknowledgement. A bounded correction-plan tool passes 14 isolated controls but grants no mutation rights or repeat-repair completion. Shared installation and the next actual image gates remain pending; host skips do not count as image acceptance. No task checkbox, live worker hold, registry, media or deployment changed.

- Next larger batch source checkpoint: the combined native route passes 22 owning controls without skips and its worker passes 46; independent final-read/lost-acknowledgement review passes five native and three worker controls. The lossless native conversion route passes 23 owning controls without skips, its worker conversion/legacy handoff passes 44, and actual archive conversion passes an additional RAR control. Parent integration passes 31 shared-writer controls and 14 private-plan controls. Scoped source snapshots (217 native files and 45 worker files) are running preliminary actual image gates. Retained-repeat reconciliation, reviewed derivative adoption and explicit conclusive cleanup are separate ongoing agent units and are excluded from those snapshots. Final integrated images, full repository checks, source delivery and live acceptance remain pending; task checkboxes and live holds remain unchanged.

- Coordinated larger-batch integration checkpoint: corrected conversion passes 26 native controls after independent review found and reproduced mutable capability redirection; the earlier preliminary native image is superseded and cannot be deployed. Retained-repeat reconciliation passes 18 owning controls plus four independent adversarial controls, preserving the correct publication and clearing only the proved false location; status and wanted-intent repair remain separately reviewed. Combined rename/metadata and cleanup pass 32 native controls; the final cleanup revision passes 13, and the worker passes 52. Read-only derivative-family validation passes 16 independent literal and actual catalog/archive vectors with no native imports. Rootless nested metadata promotion passes nine preservation controls. Startup now holds unresolved cleanup records even before a filesystem intent exists (43 host controls, seven native-source/runtime skips requiring the image gate). Shared manifests, derivative adoption integration, final matching images, full local validation and live acceptance remain in progress; none of these partial milestones completes a full remaining task.

- Final integration review checkpoint: the isolated worker candidate passed 437 tests without skips. Review then reproduced a crash before the first filesystem intent in both conversion and retained-repeat producers: an unfinished private CAS remains while ordinary startup can still admit work. A separate agent is implementing exact unfinished-record holds; matching images must be rebuilt after this change. Independent review also verified final database/marker replacement, privacy and SQLite namespace controls before ordinary admission. Combined rename/reader/metadata phases remain separate from these unfinished conversion/repeat holds. Final source delivery and live acceptance remain pending; no task is marked complete from these intermediate checks.

- Backup retention preference (2026-10-06): the user explicitly requested one verified backup and removal of older backups. Superseded operational snapshots may be removed once a complete recoverable set is independently verified. Retained media originals, active preservation receipts and unresolved correction evidence require their own acceptance and remain outside snapshot pruning. Inspection found that the newest completed rollout retained verification records after its temporary copies were removed; those records alone are not a recoverable backup. Protected-set inventory and replacement verification remain pending.

- Final source acceptance (2026-10-07): native V4 passes all 1,248 executions across 80 actual-image suites with zero skips/failures; 69 installed production modules match frozen source. Worker V6 passes all 439 tests without skips; 24 installed production modules match frozen source. Independent owning and failure reviews close T009–T012. T014 closes only the exact isolated private registration/reconciliation intent and refusal fixtures, not live preparation. Contract, quickstart and stack documentation now describe accepted source while keeping delivery, matching rollout, registration, full-library acceptance and cleanup open. Rootless automatic metadata repair is held before job creation and replay; explicit reviewed derivative publication retains its separate authority. Full local CI V5 passed before the final documentation update; final delivery validation remains required under T016.

- Delivery and snapshot retention (2026-10-07): PR254 merged as 3561ac1 after complete local CI, accepted native/worker image gates and independent reviews; branch protections were restored exactly. Compatible dependency PR255 and PR257 merged, while PR256 was closed because ComicTagger requires chardet below version 6. GitHub, local main and Gitea match 460d19a; no open PR remained at this checkpoint. A fresh independent restore check verified the retained October 4 image/configuration/worker-receipt backup (27,314 records, 64 database checks, exact manifest hash). Explicit pruning removed 154,720,058,774 bytes across 64 obsolete snapshot paths, preserving 14 currently referenced originals and all unresolved media evidence. Post-prune verification rechecked every retained backup record and all 32 captured databases. This retained set excludes worker non-JSON originals and does not replace the full consistent backup required before the next live rollout. Updated dependency image builds and complete T017–T019 operational acceptance remain pending.

- Parallel DDL correction checkpoint (2026-10-07): separate source owners implemented completed-pack recovery and DDL-to-NZB removal, with independent reviews before parent integration. Pack recovery accepts filename volume ordinals only after an exact unique catalog/version match, translates cleaned target paths through explicit native/worker mappings, and requires current owner/archive/full-hash proof before report refresh; failed rechecks retain historical member proof. The owning actual-runtime suite passes 33 tests without skips, with 35 import and 13 native handoff controls also passing. Read-only checks of the retained Grimm and sample completed-pack records verify 137/137 current destination hashes, signatures, ZIP CRCs and exact catalog owners. Six later Grimm captures still require fresh inventory and guarded reconciliation; these checks do not claim all DDL captures are complete.
- DDL-to-NZB removal checkpoint (2026-10-07): native Remove previously rejected every reserved handoff although the queue offered the action. The reviewed change releases an exact queued reservation and conditionally deletes only its unchanged native DDL row; stale search tokens cannot submit afterward. Accepted downstream jobs retain their ownership and processing; uncertain in-flight states remain reviewable. Immutable removal receipts support lost-acknowledgement reconciliation and reject reused IDs, changed row provenance or ownership. Isolated replay of all 65 actual queued handoff projections passes; no live DDL row, NZB job or media file was removed. Combined native V6/worker V7 actual-image gates and repository validation are running; consistent full-state backup, verified rollout and live acceptance remain pending.
- Dependency runtime correction checkpoint (2026-10-07): offline CLI validation found Acquire's pinned Dissect runtime cannot start under Python 3.14. The reviewed repair pins the official Python 3.13 base, preserves package versions and hashes, and adds an actual CLI-startup build gate. Exact lock regeneration, dependency consistency, CLI startup, flow-record serialization and forensic virtual-filesystem reads pass in the rebuilt image. Five other updated CLI images pass isolated checks; live deployment refresh is not yet claimed.

- Combined-image checkpoint (2026-10-07): the final worker V7 passes 444 actual-runtime tests without skips or failures, and all 24 installed production modules match frozen source. Full local CI and history secret scanning pass (528 commits). Independent integration review clears the Acquire runtime repair and the DDL/pack documentation against the implementation. The all-27-completed-DDL diagnostic refuses its final proof because the running native catalog changed during the read; no report or completion claim was written. Repeat this audit against a fresh quiesced baseline during the backup window. Native V6 reliability acceptance, source delivery and verified live recovery remain pending.

- Final DDL image and delivery acceptance (2026-10-07): native V6 passes 1,259 actual-image test executions across 80 suites without skips or failures; all 69 checked installed production modules match frozen source. Worker V7 passes 444 tests without skips or failures with exact 24-module parity. PR258 merged as 9b3690b after independent review and full local validation, using the user’s hosted-CI waiver; branch protection was restored exactly. GitHub and Gitea main match the reviewed merge. The primary local checkout remains at its prior revision until backup source verification completes, because its Mylar stack files are part of the consistent backup. Source delivery does not complete live recovery or the remaining task checkboxes.

- Operational hold and backup verification checkpoint (2026-10-07): Mylar is stopped and the worker remains held while the consistent full-state backup runs. Its 57 explicit scopes include native state and retained worker originals. Independent review reproduced a late source-ancestor alias gap in the original verifier; a compatible corrected verifier passes 41 controls in the actual native runtime, including the reproductions. The new full backup must pass this corrected independent restore/source verification and a fresh container-state check before rollout. The prior verified retained backup remains protected until that acceptance. The user is away and has suspended new pkexec commands; no restart, deployment or protected audit is attempted during this hold. A pending read-only DDL export was cancelled before Docker started; repeat it against a fresh quiesced snapshot when access returns.

- Dependency deployment checkpoint (2026-10-07): Acquire, Blackbird and Metagoofil selectors now reference the accepted local images after configuration backup/restore and preservation checks; their absent containers were not started. Privotron was refreshed to its accepted image while preserving its Created state, command, configuration, networks and empty profile volume, with independent preservation checks. The Torbot pair update remains unexecuted: independent review clears 23 isolated lifecycle/preservation controls, while actual Docker cloning, archive attributes, retained Tor state and health still require live acceptance. Full-library, registration, completed-DDL recovery and final cleanup gates remain open.

- Backup copy completion (2026-10-07): the already-running consistent full-state backup command exited successfully with Mylar stopped and the worker held. This establishes copy completion only. Corrected independent restore/source verification and a fresh protected container-state check have not run; rollout and pruning remain held. No new privileged command was issued while the user is away.

- Consistent full-backup acceptance (2026-10-07): after privileged access resumed, corrected independent verification passed against the existing full backup and detached restore, preserving 30,616 files, 7,271 directories, one symlink and 38 databases. Fresh protected post-verification checks prove the original Mylar/worker identities, stopped/Created states, configuration, mounts and static networks unchanged. The primary local checkout now fast-forwards to reviewed GitHub/Gitea main at 4090ced. Native-first image rollout, registration and current-library acceptance remain open; the verified backup remains retained.

- Fresh completed-DDL audit (2026-10-07): the exporter now checks the actual source-verified 19-column schema and excludes download-link fields. All 20 exporter and 18 diagnostic controls pass without skips, with independent receipt/namespace review. The actual quiesced export and full selected-row diagnostic pass for 26 current Completed rows: seven pack rows have 26 retained captures, eight members retain matching destination hashes and owners, and 25 captures still need inventory. Of 19 single rows, 16 have a proved current catalog/file owner and three require further review. These are read-only facts, not import acknowledgements or cleanup authority. A private targeted-role backup adapter passes 27 controls and independent late-change reproductions against both production consumers; actual registration still needs fresh scoped source/census evidence and native acceptance.

- Native-first rollout checkpoint (2026-10-07): the running Mylar container selects the accepted native V6 image and all 69 installed production modules match the accepted image. Authenticated passive publication status reports held/authority-unavailable. The worker container selects accepted V7, remains Created, and preserves its mounts. Ordinary native operations remain blocked until successful initialization and a later restart; source review and a disposable fixture confirm that bootstrap and registration do not release this startup hold. Current catalog preservation, actual authority initialization, correction registration, completed-DDL recovery and full-library acceptance remain pending.

- Authority and catalog acceptance checkpoint (2026-10-07): read-only immutable checks prove all 24 live catalog tables, 56,136 typed rows and the complete schema match the accepted detached restore, with no changed tables and readable databases. The primary-key protocol prepared and committed the first publication authority using the actual verified manifest and restore-result hashes. Authenticated post-commit health confirms the same native process remains held with startup-restart-required; the worker remains Created. No media recovery, correction-family registration or DDL cleanup has occurred. Fresh role proofs must support this verified startup-held process without restarting before protective registrations; Python 3.10 compatibility of the private role adapter is separately verified. T017–T019 remain open.

- T016 delivery acceptance (2026-10-07): reviewed source and documentation PR258/PR259 are merged; full local CI and final independent reviews passed under the user’s GitHub-CI waiver, with branch protection restored exactly. Fresh remote checks show GitHub, Gitea and primary local main all at 4090ced, with no open PRs. Actual selected native V6 and held worker V7 image identities match their independently accepted image/module gates. This closes delivery; operational registration, current-library recovery, restart/canary acceptance and cleanup remain under T017–T019.

- Live registration checkpoint (2026-10-07): five Heavy Metal payload families have actual native registrations at successive census revisions 1–5. Each uses fresh exact owner/source facts, a private four-role backup, independent detached restore and post-copy proof that the same native process remains startup-held. Correct 1977 archives remain unchanged; the five 2025 rows are Snatched with no Location claim, so the owned-repeat repair route is inapplicable and no status, wanted intent or media file is changed. Both the worker hold and startup-restart-required health remain verified. Additional historical identity families, current-library acceptance and live processing/restart canaries still require fresh proof before release.

- Additional identity registration checkpoint (2026-10-07): eight historical payload families covering Batman/Scooby-Doo, Creature from the Black Lagoon and Guardians have native registrations at census revisions 6–13. Each registration binds fresh current corrected ownership, exact retained member evidence, its own four-role backup and independent restore, and verified post-copy startup exclusion. The rejected tuples are payload-specific: Guardians issue 507776 remains allowed for its legitimate different payload. Actual final health confirms all 13 registered families, the same startup-held native process and the Created worker. Registrations changed no media, catalog ownership, acquisition status or wanted intent. Ten earlier Heavy Metal advisory checks passed; additional advisory canaries, processing/restart checks, complete current-library acceptance and DDL recovery remain open.

- Additional advisory canary acceptance (2026-10-07): 17 actual primary-key checks against the final census revision 13 allow all eight reviewed corrected owners, hold all eight wrong payload/owner tuples, and separately allow the legitimate different Guardians payload. Every response binds the same complete census and exact native process identity; the reviewed helper passes 26 host and selected Python 3.10 runtime controls without skips. This proves advisory decisions only. It does not complete media processing, full-library recovery, restart acceptance or general worker release. Full repository validation and local CI pass for the registration checkpoint, with no secrets detected.

- Current-library verification checkpoint (2026-10-07): the selected native image's read-only audit is running across 3,900 discovered archives under the existing writer exclusion. Its helper passes 33 host and actual Python 3.10 image controls without skips, verifies all 69 installed SDK modules, and records complete per-archive member/page inventories. Progress counters are preliminary; final source/namespace/authority continuity and post-process state acceptance remain pending. Mylar stays startup-held and the worker stays Created. No rename, metadata write, DDL recovery or startup release is inferred from this scan.

- Startup-held DDL inspection readiness (2026-10-07): the truthful held-process exporter passes 28 host and native-runtime controls, while the diagnostic passes 47, with independent review of late SQLite sidecar and output changes. These are helper readiness checks only. A fresh export must wait for the current audit to release writer exclusion and must bind actual process, census, parity and post-process proof before any recovery decision. T017–T019 remain open.

- Recovery verification checkpoint (2026-10-07): all 3,900 current archives have reached inventory, while final native owner, full-source and namespace continuity checks remain pending. The accepted-audit supplement comparator passes 32 host and selected native-runtime controls, and the held DDL parent passes 22; neither has used pending audit output as acceptance. A corrected query-only reader observer passes 18 host and actual Python 3.10 controls with independent review, and the actual image/application JAR matches the pinned Komga 1.28.1 source contract. Reader absence, unowned-library retirement and publication remain explicitly unproved. The independent review of a separate all-catalog absence component closes an initial ancestor gap, but that component grants no Writer, census, current-owner or retirement rights. T017–T019 remain open.

- Dependency verification checkpoint (2026-10-07): the GET-free disposable Tor filesystem proof passes native preservation and strict provider continuity checks; independent checks confirm both disposable containers are absent. The proof establishes only a verification platform and cannot substitute for a fresh actual candidate proof. Before the new rollout, the live pair is verified healthy on its prior images. The revised updater still requires selected-runtime checks, fresh candidate acceptance and live preservation/health verification before its dependency update can be considered complete.

- Live reader-reference checkpoint (2026-10-07): the corrected query-only observer verifies two complete stable API observations covering 3,927 books, 584 series and two libraries, with fresh matching process and application-JAR proofs. All five rejected library paths and their five correct counterparts have reader entries; one rejected path also retains a deleted duplicate entry. These actual references require a separate preservation/transfer decision before retiring the wrong files. The query grants no reader absence, catalog ownership, media mutation or publication authority. Current-library recovery and final-release tasks remain open.

- Tor rollout rollback checkpoint (2026-10-07): the revised updater verifies the actual fresh candidate filesystem and reaches healthy candidates, then refuses the final rendered Compose preservation comparison. A separate read-only post-check confirms the original pair, selectors, protected data and health are restored, with the candidate and verifier containers absent. The difference is the second image selector rendered into the application service through its existing environment file; the comparison expected that selector only in the Tor service. Operation evidence and backups remain retained while this exact projection is corrected. The dependency update is not accepted.

- Tor updated-image checkpoint (2026-10-07): the exact selector-projection correction passes independent review, 52 updater controls in the selected runtime and 19 candidate-verifier controls, without skips. The subsequent scoped update completes its fresh backup/two restores, actual candidate filesystem proof, healthy startup, protected-data comparison and full Compose preservation check, and removes only its own completed temporary copies. The application remains idle. A separate read-only post-update check also passes: both exact images and expected processes are healthy, configuration and all eight native Tor state members are preserved, prior containers are absent, and the current labeled verifier inventory is empty. The latest retired verifier exact ID is unavailable after producer cleanup, so the independent check records that limit rather than reconstructing it. Prior rollback evidence and referenced backups remain retained while comic recovery is unresolved.

- DDL recovery batch readiness (2026-10-07): the detached-evidence recovery planner passes 23 host and selected Python 3.10 runtime controls, including independent reproduction of an initial ancestor-replacement gap. It joins the exporter, diagnostic and parent terminal proofs before classifying the 26 selected rows for guarded status refresh, import review or pack-member reconciliation. It performs no API, catalog or media operation and grants no import or cleanup acknowledgement. Actual planning waits for a fresh accepted export after the current library audit releases exclusion; execution additionally requires accepted native release canaries.

- Reader correction preparation (2026-10-07): separate schema-observation and full reader-state backup helpers pass 30 and 26 selected native-runtime controls without skips. The protected inspection runtime also passes its 30 isolated controls and Docker compatibility check. The reviewed outer launcher has 27 host controls; its matching runtime gate and actual live schema observation remain pending. The earlier general rollout snapshot contains reader stack configuration, but does not cover the reader databases. A separate consistent reader-state backup and isolated restore are required before correcting the five active rejected book entries. Preserve all six rejected book IDs, their references, and the five correct counterparts. No reader or archive correction has run from this preparation.

- Negative-retirement readiness (2026-10-07): independent review reproduced and closed two private journal gaps: a platform callback could change a source before a late refusal, and proof expiry could strand recovery. The corrected core passes 31 local controls without skips, keeps changed files in place, and requires explicit durable renewal for fresh recovery authority. It grants no native publication or import acknowledgement. The actual library uses NFS, which the local-only core refuses; selected-runtime checks, actual NFS no-overwrite behavior, native authority integration and the separate reader-preservation transaction remain unverified. No rejected archive has been retired.

- Audit and recovery checkpoint (2026-10-07): the read-only whole-library audit has completed source, namespace and held-process verification for 3,900 archives. Seven archives remain unavailable to inventory, with 26 additional metadata previews unavailable; the audit grants no publication acceptance. A fresh held-process completed-DDL export is in progress after audit writer exclusion released. The negative-retirement journal and filesystem-probe helpers pass 49 actual native-runtime controls without skips, but actual NFS behavior, native authority integration and reader preservation remain unverified. Mylar remains startup-held and the maintenance worker remains Created; no archive retirement, reader correction or DDL recovery is claimed.

- Reader inspection checkpoint (2026-10-07): 62 isolated outer-inspection controls pass in the actual selected runtime without skips. The subsequent protected observation correctly refuses an unexpected anonymous writable image volume before starting its database observer. A narrow correction explicitly covers that destination with the same approved read-only configuration source and passes 36 host controls; matching runtime checks and a new actual observation remain pending. Fresh completed-DDL exports verify the held process, but diagnostic acceptance remains unproved. No reader, archive or DDL mutation has run from these checks.

- Reader schema acceptance (2026-10-07): the corrected inspection profile passes 70 isolated native-runtime controls, and the actual read-only schema observation verifies both reader database schemas against the unchanged live process. The implicit writable image config volume is explicitly shadowed by the existing read-only config bind. The five-entry correction kernel passes 34 host controls and independent review using isolated synthetic rows; live application is disabled. A fresh observation finds mapped index segments under the reader config directory, but complete effective index configuration, a consistent full reader backup/isolated restore and native file-retirement proof remain required. No reader correction, archive retirement or publication acceptance is claimed.

- Recovery diagnostic progress (2026-10-07): the targeted unavailable-archive diagnostic passes 28 actual native-runtime controls, with execution against the seven inventory failures and 26 additional metadata-preview failures awaiting its checked parent. The completed-DDL diagnostic and parent pass 71 actual native-runtime controls; a fresh read-only export is captured, and the revised failure path retains bounded private error details when the diagnostic refuses execution. Neither prior refused attempts nor fresh export capture establish recovery or import acceptance. All live publication holds remain in place.

- Completed-DDL diagnostic acceptance (2026-10-07): the actual diagnostic refuses worker receipt paths outside its declared scope. The checked launcher now verifies the existing worker cache mapping and exposes that same cache read-only at the worker's recorded path; it passes 25 host controls and independent review, plus 73 actual native-runtime diagnostic/parent controls. A fresh export and diagnostic then verify all 26 selected rows with the same held process. The detached recovery planner passes 23 selected-runtime controls and classifies six packs for ordinary member reconciliation, one pack for retained-source review and nineteen singles for guarded import review. Planning grants no import, queue removal or cleanup acknowledgement. The targeted archive/metadata failure observer is running after 32 parent controls pass in the selected runtime and independent review; final library acceptance and release remain open.

- Fresh recovery evidence (2026-10-07): the targeted observer verifies all 33 selected archive/metadata failure paths with unchanged source and held-process bindings. All 26 metadata previews are available in this fresh observation; the earlier suppressed exceptions do not establish their original cause. Seven native inventory refusals remain: four archive-decoder CRC errors, one combined ZIP-member safety refusal and two PDFs routed through the generic TAR decoder. These are diagnostic results, not corruption proof or conversion authority. A revised detached DDL planner passes 28 actual native-runtime controls and classifies seven packs for retained-source inventory, sixteen singles with historical downloaded-owner/file evidence and three unproved singles. One missing older pack capture remains separately held; no cleanup or import acknowledgement is granted. Final current-library recovery and release remain open.

- Reader and archive diagnostic progress (2026-10-07): the effective-scope observer now covers the actual fourth reader mount, stable process identity and the four packaged configuration resources at the pinned JAR root. A read-only check confirms the optional fonts directory is absent. The observer records that absence without creating a directory and passes 63 isolated native-runtime controls with independent review; fresh live scope acceptance remains pending. The corrected seven-file diagnostic passes 98 isolated controls and completes its actual read-only checks with source and held-process continuity. All four RARs pass the independent unrar integrity test but still fail the native decoder with member CRC errors, so corruption and native compatibility remain unproved. Both PDFs render every page successfully, while their native archive route remains unsupported. Lady Death has a zero-byte directory entry missing its trailing slash; regular members pass CRC checks, but no repair or publication is inferred. The reviewed prior checkpoint is merged and GitHub/local/Gitea main match. Reader backup, archive recovery, DDL reconciliation and final publication acceptance remain open.

- RAR decoder correction preparation (2026-10-07): all four refused RAR4 archives stream successfully through the unchanged verifier on host libarchive 3.8.9, with complete before/after source hashes and file identities unchanged. This is diagnostic comparison only, not selected-image or publication acceptance. The reviewed candidate builds official hash-pinned libarchive 3.8.9 on the existing base, runs upstream tests, verifies the loaded library and retains full CRC checks. Six host controls and independent source review pass. The isolated candidate build completes successfully, including upstream tests and the existing native image gate. Image identity, actual ELF/dependency checks, the four retained RARs in that image, matching worker verification and restore-verified native rollout remain pending. Reader-owned CBR conversion also remains held until a supported transaction preserves scoped reader references and sidecars before native catalog adoption. Existing startup and worker holds remain in place.

- Reader verification checkpoint (2026-10-07): the fresh effective-scope observation is accepted with unchanged reader process identity after 64 isolated controls pass. The separate NFS readiness gate passes 58 isolated controls; no live canary, archive retirement or publication is authorized by that receipt. The restore-row gate exposes one selected-runtime failure among 27 controls, and independent backup-parent review reproduces acknowledgement after restart-policy drift. Both helpers remain held for correction and renewed verification. Komga has not been stopped for the new backup, and Mylar and its worker remain held. No task completion is inferred from these diagnostic results.

- Authenticated diagnostic checkpoint (2026-10-08): decoder source PR #265 is merged and GitHub/local/Gitea main match. The stored candidate image loads libarchive 3.8.9 from the intended opt library; actual runtime verification checks all 69 installed SDK modules, and the corrected verifier/probe passes 31 isolated controls. This does not verify the four retained archives or authorize rollout. Restore-row observation passes 27 actual runtime controls. The backup-parent fixture passes 37 actual runtime controls after a test-capacity correction preserves the live disk-space guard. The first live backup attempt stops Komga but refuses Docker's deprecated stop-flag warning on stdout before copying state; the unchanged reader is restarted. The reviewed replacement uses the supported timeout flag and retains strict acknowledgement checks. A fresh reader scope, corrected runtime gate, full backup/restore acceptance and final library/DDL recovery remain required. Existing task checkboxes remain open.

- Reader backup correction verification (2026-10-08): the corrected stop helper retains exact acknowledgement checks, uses the supported timeout flag and preserves private command diagnostics. Independent review and all 41 isolated controls in the selected runtime pass. A fresh reader scope observation verifies the restarted unchanged process; the previous authentication-delayed observation remains unaccepted. The new full configuration backup and independent restore have reached isolated container creation after the exact stop acknowledgement passed. The first attempt copied no data and remains retained as failure evidence. Actual backup acceptance, reader correction, archive recovery and service release remain pending; no task completion is claimed.

- Reader backup acceptance (2026-10-08): the full persistent configuration backup and independent restore pass completeness, file integrity, database readability and all eleven targeted reader-row checks. The stopped source and both copies retain their verified bindings. The unchanged reader then starts successfully and a separate check confirms it is healthy; its new process makes earlier scope observations historical for subsequent lifecycle actions. The verified copies remain retained for the unfinished correction. No archive retirement, reader SQL update, DDL mutation or writer release is claimed.

- Malformed ZIP verification preparation (2026-10-08): the read-only exceptional-directory inventory producer passes independent review and all twenty controls in the selected native runtime, with all 69 installed SDK modules verified. Its first runtime harness collected zero tests and was rejected; the corrected harness explicitly collects the complete suite. No actual ZIP repair, derivative equivalence or publication permission is inferred. The four-RAR parent and native negative-purpose observation modules remain in review; a reproduced late catalog-claim gap keeps the latter unaccepted until its final observation lifetime is corrected.

- Decoder and preservation controls (2026-10-08): the four-RAR parent passes 23 selected-runtime controls without skips, but its actual observation stops before decoding because Docker prepends a newly supplied environment variable. The next correction preserves exact environment values while matching that ordering. The reader disk adapter passes 26 disposable database controls; operational admission remains disabled. Independent review also finds a late journal-entry race in the proposed hardlink transition, which requires correction before runtime or filesystem use. These results do not establish archive recovery, reader correction or publication acceptance.

- Selected-image decoder observation (2026-10-08): all four RARs complete streaming inventories with 28, 186, 28 and 60 members using the candidate libarchive, and independent archive tests exit successfully. The checker acknowledgement verifies its report, but the outer parent stops on a library-report shape comparison before final acceptance; the disposable reader is stopped unchanged. A later disposable rollback test also finds that the proposed reader disk adapter can overwrite a foreign database change made during custody verification. Both helper corrections and their runtime checks remain required; no recovery, retirement or publication completion is claimed.

- Four-RAR compatibility acceptance (2026-10-08): the corrected outer parent passes 26 selected-runtime controls and completes the actual same-container observation of all four original archives. Full inventories, native streaming, independent archive tests, unchanged source identities, loaded library and installed module parity pass; the test container is retained stopped. This verifies candidate decoder compatibility only. Deployment, reader correction, full-library publication acceptance and DDL recovery remain open. Repository validation passes.

- Correction runtime controls (2026-10-08): the revised reader disk adapter closes late current-state, restore-copy and rollback-journal changes. Its 37 controls and 24 native negative-preparation controls pass together in the selected Python runtime with installed module parity and no skips. These disposable tests do not enable operational admission. Rollout review also establishes that the existing startup hold is process-local, so recreating Mylar could start queues; the verified decoder remains staged while complete library acceptance and the owning recovery integration remain open.

- Built-in archive repair preparation (2026-10-08): the exceptional ZIP inventory and lossless derivative pass 37 isolated native-runtime controls. The corrected diagnostic parent passes 29 controls, and the combined 66-control selected-runtime gate verifies installed module parity with no skips. A fresh read-only header capture identifies one empty directory missing its trailing slash; actual private-copy verification is pending. A separate subagent packages the repair classifier and derivative with 21 host controls. CRC failures, truncation, encryption, unsafe names and ambiguous members remain held; RAR decoder compatibility and PDF conversion retain separate paths. These helpers do not grant publication or replace originals. Independent package review, native package tests and owning recovery integration remain required.

- Built-in repair runtime verification (2026-10-08): independent byte-fixture review confirms that changed payloads, CRC failures, truncation and altered witnesses are refused. All 21 package controls pass in the selected Mylar Python runtime with installed module parity and no skips. The actual malformed archive diagnostic starts its isolated container but stops at inventory production before writing a repaired copy; read-only checks confirm the original hash and file identity still match. Operational adoption and publication remain disabled while that refusal is diagnosed.

- Real ZIP diagnostic refusal (2026-10-08): read-only tracing identifies a manifest permission mismatch before archive inventory: the helper requires a private SDK map, while the diagnostic mounted the unchanged public map with mode 0644. The original full hash and file identity match, and no derivative was written. A revised parent is being prepared to create an exclusive private map copy with bound integrity checks; neither shared source permissions nor protected evidence requirements are weakened.

- Archive diagnostic review hold (2026-10-08): the private-map correction passes 69 isolated native-runtime controls, but independent review and a reproduced disposable fault show that final result collection can mask a map change after container stop. The corrected parent has not been used on the original archive. A further revision must preserve the initially captured map binding and pass late-change regressions before actual use; the prior failed diagnostic container is retained stopped with no repaired output.

- Original ZIP inventory verification (2026-10-08): the corrected parent passes independent review and all 73 selected-runtime controls. Read-only inventory production then succeeds on the original archive, including complete member CRC checks. The separate derivative step refuses a missing metadata-module import in the isolated launcher before any output write. The next launcher revision must load the installed SDK through its verified import helper and exercise fresh-process metadata handling; no repaired library file or publication acceptance is claimed.

- Lossless original ZIP diagnostic acceptance (2026-10-08): after independent review and 77 selected-runtime controls, the revised parent verifies a private derivative of the original malformed ZIP. All 30 members and 28 ordered pages retain identical payload and root metadata; full CRC checks pass before and after. Only two slash bytes and the required ZIP name-length/offset fields change, increasing the archive from 27,782,320 to 27,782,322 bytes. The original full hash and file identity are unchanged, and the isolated container is retained stopped after exact acknowledgement and terminal checks. This proves the bounded repair as a private diagnostic. Automatic adoption, reader/native publication and library replacement remain pending in the owning recovery integration.

- Archive recovery implementation dependency (2026-10-08): [feature 024](../024-verified-archive-repair/tasks.md) defines the separately owned lossless repair route. Its ordered source, runtime, reader/native adoption and live acceptance tasks preserve these existing task IDs and release gates. Diagnostic integration proceeds in parallel with native negative-publication recovery; private repair proof does not complete this feature.

- Reader recovery prerequisite verification (2026-10-08): independent review closes the original native-baseline resealing and late reader-pair adoption faults. All 90 coordinator, reader-admission and native-negative preparation controls pass in the selected Mylar Python runtime with installed SDK parity and no skips. These isolated tests grant no live execution. The operational reader/filesystem commit and rollback capability remains unfinished, so live library, DDL and writer-release acceptance remain open.

- Diagnostic delivery and owning recovery scope (2026-10-08): PR #267 delivers independently reviewed archive diagnostics and supplement descriptor checks after local CI, a 1,317-control native baseline, 42 final affected native controls and 500 worker controls without skips. GitHub/local main and the authorized mirror match the reviewed squash. The separate negative-retirement prototype has 41 phase, 43 preparation and 16 reader-projection host controls, independently reviewed as source-only. Its aggregate factory, shared-parent projection, reader SQL custody and terminal closure remain unfinished under T020–T027. No live writer release or recovery acceptance is inferred.

- Owning recovery source integration (2026-10-08): primary-key archive preparation/status and native/worker negative-retirement purpose denial are integrated as the next source batch. Preparation retains separate original/restore custody and every adoption/publication/reader grant remains false. Existing release gates stay open while independent integrated review and actual installed-runtime checks run. The five-source aggregate projection correction binds journal output to its admitted directory descriptor; its owning SQL/rollback/terminal consumer remains under implementation. No live library/catalog/DDL or writer-release change ran.

- Aggregate terminal implementation review (2026-10-08): marker removal before durable completion creates a crash window that can admit ordinary processing. T025 now explicitly requires a root-owned durable successor hold across marker unlink, fsync and terminal acknowledgement, including interrupted/restarted admission controls. SQL commit/reversal, typed handoff and terminal implementations are running as separate coordinated source proposals. No prototype or source-only test grants live execution.

- Owned preparation image acceptance (2026-10-08): 298 selected native controls pass without skips in the corrected installed image, including genuine Controller/Writer preparation and all affected tagging controls. The full worker image passes 504 controls without skips and its final 28-module non-root smoke passes. T020–T027 aggregate execution prerequisites and T017–T019 live acceptance remain unfinished. No reader/native database, library, DDL or service mutation ran.

- Terminal admission source verification (2026-10-08): native and worker Writer admission now also refuses the fixed terminal successor hold, including its appearance after the final semantic fence callback. Archive preparation/status perform the same final direct refusal. All 130 focused host controls pass without skips. The separately reviewed SQL commit, terminal completion and rollback prototypes remain source-only. T020–T027 remain open until integrated installed-runtime and operational evidence passes. Privileged operations are paused at the user's request. No live state or service changed.

- Recovery implementation progress (2026-10-09): the source adds both archive-repair pending/terminal markers to native and worker Writer admission and the final direct preparation/status checks. All 108 affected host controls pass without skips, including malformed markers, recovery-flag refusal and markers introduced by the last callback. Independent review repeats both marker names after each last callback and confirms refusal. The selected Mylar Python 3.10 runtime passes all 99 native controls without skips in an isolated container using public source and disposable fixtures; this is runtime compatibility proof, not installation or live recovery acceptance. The live Mylar source check confirms its explicit DATA directory before configured-root binding; no service or library state changed. Reader admission, WAL commit/rollback and parent/provider integration remain under review. No live recovery, DDL import or broad library acceptance is inferred.

- Reader and dispatch integration progress (2026-10-09): PR #271 delivers ordinary Writer refusal for both archive-repair holds; all three main copies match `86d9c912`. Independent review clears the current admission, reader phase, SQL commit and aggregate sources after 168 host controls and 21 composed WAL controls. A new provider has 16 boundary controls and corrected native-to-host-to-reader geometry has 45 lifecycle controls. The current tracked integration has 62 archive adoption, 30 worker transport and 39 owning API controls. Review found and corrected a queue API call made before Writer construction; both request and status now acquire the actual existing Writer. Standalone terminal reconstruction, partial-execution rollback orchestration, installed factories, continuous parent and actual NFS acceptance remain unfinished. A read-only authenticated native health observation confirms startup-restart-required and census revision 13. T017–T019 and T020–T027 remain open; no reader/catalog/library mutation, DDL import or queue cleanup occurred.

- Integration follow-up (2026-10-09): the installed source list now includes the checked reader lifecycle and configured-native-scope modules. Their portable fixture suites pass 45 and 28 host controls without skips; source-pin substitutions in the latter are explicit authored fixtures, not actual installed factory acceptance. Review also found periodic repair dispatch starving jobs beyond its first 32 and aborting unrelated maintenance after HTTP failure. Round-robin selection and transport-only exception containment fix both; the 30-control worker suite covers 33 jobs across two cycles and lost-response retention while another job proceeds. Owning imports, installation parity, genuine parent/provider execution and live acceptance remain pending. Task ordering and existing completion requirements are unchanged.

- Integration runtime check (2026-10-09): all 204 selected controls pass without skips on the actual Mylar Python runtime: 62 archive adoption, 39 owning API, 45 lifecycle, 28 configured-scope and 30 worker transport checks. This isolated run mounts frozen public source only and uses disposable fixtures; it does not prove candidate-image installation or live recovery. Repository validation, Ruff and full local CI pass. Independent review verifies queue request/status with the real fixture Controller/Writer, all 33 jobs reached across two dispatch cycles, and timeout retention without blocking another job. Original task completion gates remain in force.

- Recovery follow-up (2026-10-09): independent source review finds four additional lifetime/mapping defects: a parent callback can substitute a native proof, configured scope can refresh mutable seals, a reader destination overmount can shadow the native source, and the inherited pipe can be redirected by refreshing descriptor facts. The corrective source anchors original seals and pipe identities, snapshots proof before callbacks and checks reader destination-to-host projection. All 52 lifecycle and 31 configured-scope host controls pass; the original four reproductions now refuse. The pinned upstream argparse source also accepts abbreviated configuration/datadir switches; finite exact launch-option parsing refuses those aliases. Corrected-image and delivery gates are still pending.

- Worker and rollback progress (2026-10-09): the frozen PR #272 worker candidate passes all 540 actual image controls without skips. Its final non-root smoke verifies all 29 production module bytes against frozen source and zero effective capabilities. A separate source-only typed rollback provider passes 29 host controls, including actual SQLite five-row reversal and five-source filesystem restoration; lost COMMIT/response, closed WAL connections and uncertain phases retain custody. It is not installed. The native baseline image now passes 1,585 controls across 90 suites without skips. A corrected image layer passes 84 selected controls (52 lifecycle, 31 configured-scope and one installed archive-factory check); its final image copies only the two corrected production helpers. A read-only final-image check confirms exact source bytes, absence of test files, UID 1000 and zero effective capabilities. Full local CI passes, including history secret scanning. These checks do not establish genuine stopped-reader parent execution or live recovery acceptance. Continuous parent, durable independent terminal observation, actual NFS, library/DDL recovery and release remain unfinished. No live service or library state changed.

- Recovery delivery and portable-test progress (2026-10-09): PR #273 is merged at `7677bae8`; primary local main, GitHub and the authorized mirror match. A separate canonical recovery-cohort proposal now has 682 host controls across 22 suites, with zero skips; root repeats the complete run from a relocated copy. Tests bind the exact successor source bytes and keep predecessor fixtures outside the production installer. These are source controls, not installed factories or live recovery acceptance. Parent handoff, independent terminal observation, actual NFS behavior and final library/DDL acceptance remain open.

- Canonical recovery integration (2026-10-09): the tracked adapter installs 21 new helpers and checks three existing helpers, with exact pins for all 24. The private-copy build runner corrects checkout file-mode and root-build fixture ownership without changing production bytes; root repeats all 682 host controls and the actual native layer passes the same 682 controls plus fresh installed origins/hashes/SDK identities. The final-image check verifies all 24 production source bytes, excludes tests/fixtures, and confirms UID 1000 with no effective capabilities. The provider/preimage/terminal-observer suite passes all 59 controls twice in actual Mylar Python 3.10.12, including a moved-tree repeat; its native/lifecycle boundaries remain explicit fixtures. Full staged local CI passes. Ruff finds and fixes one redundant test import; the affected 47 host controls and full changed-source Ruff check pass afterward. Source review identifies historical backup/row producer pins that must be updated coherently with fresh nine-role proof production before owning admission; parent pin remains absent. No live recovery or release is claimed.

- Coherent reader proof checkpoint (2026-10-09): PR #274 is merged at `d29a956b`; primary and integration checkouts, GitHub and the authorized Gitea mirror match. The corrected nine-role backup producer and admission consumer preserve the current canonical negative/lifecycle source pins and use actual backup/row producer provenance. All 75 portable controls pass twice in Mylar Python 3.10.12 at UID 1000 with zero effective capabilities and no skips, including a moved-tree repeat. These isolated source fixtures do not prove connected installed factories, fresh stopped-reader backups or live recovery. The new running-child metadata helper and exact lifecycle successor pass 104 additional selected-runtime controls (21 birth, 52 lifecycle, 31 scope), with no skips and real OS pipes against an explicit fixture parent. This is not a real selected provider handoff. The parent successor fixes a harmless publication-counter false hold, but root and independent review then reproduce a late admitted-source change accepted after the final selected-child inspection. V5 remains rejected. The V6 successor carries every original source/proof vector into the final raw closure; all 52 owning controls pass in the selected Python 3.10 runtime. Independent review repeats the original failure plus six late source/proof/ancestor changes, all refused with the acceptance flag false. Its actual protected parent invocation and native producer remain unverified. Parent authorization remains disabled; T022/T026/T027 and live acceptance remain unchecked.

- Connected source integration checkpoint (2026-10-09): integrated the canonical scope-birth helper/lifecycle successor, current backup/rows admission pins, provider same-token handoff, nine-role backup producer and independent terminal observer. The public recovery tools remain outside the final runtime package. Review reproduces an installer first-parse metadata refresh; the correction passes a red/green regression and independent literal callback checks, with all 13 installer controls passing. Git-mode integration initially fails the preimage controls because public sources are 0644; the corrected runner copies only manifest-declared fixtures into a private 0600 tree and preserves original bytes/permissions/incarnations. All 721 cohort controls, 137 recovery test executions and eight runner controls now pass on the host, without skips; repeated admission/runner cases are not counted as distinct acceptance. The corrected frozen image build passes all 721 cohort, 137 recovery and eight runner test executions in Mylar Python 3.10.12, and installed parity verifies all 25 canonical modules. Full staged local CI passes after correcting an executable source-file mode. The final actual factory handoff then exposes a missing bundled dependency search path in the standalone provider (`bencode`); owning loader correction and repeat final-image proof are pending. Parent V8/native-observation adapters and protected handoff remain separate unfinished work. Current reader-pair comparison already checks proposed signatures against the actual child kernel; positive equality on this deployment remains to be verified rather than assumed. No source/image test releases the live hold.

- Installed handoff follow-up (2026-10-09): actual canonical SDK loading exposes the missing fixed bundled-library bootstrap and then the pinned upstream image’s group-writable config/main sources. The provider now uses the fixed bundled path, and the candidate build closes config/main modes to 0644/0755 while preserving their exact bytes; fresh image verification checks those source boundaries. All 721 cohort controls, 143 recovery executions and eight runner controls pass in the selected runtime after the bootstrap correction. Independent SDK review then reproduces an original ancestor being refreshed between source reads; the corrected loader retains first map/source ancestors and refuses conflicting later merges. All 33 provider controls pass, and independent review repeats the original refusal. Native observer V2 passes 34 host controls and peer review of the corrected launch/process closure, but actual protected parent/daemon and full live acceptance remain unfinished. No live service, library or database mutation ran.

- Final connected-image verification (2026-10-09): the corrected final image passes 721 cohort, 146 recovery and eight runner controls with zero skips in Mylar Python 3.10.12. Fresh verification checks all 25 installed production module bytes, exact canonical SDK/type sharing, config/main pinned bytes and safe modes, UID 1000 and zero effective capabilities. Genuine installed birth/lifecycle/configured-scope factories share the same original pipe through eight challenge frames; parent observations remain explicitly synthetic and every live/publication grant stays false. The owning SDK ancestor regression and independent review pass. Full staged local CI passes. Protected parent invocation, current backup/restore, NFS and complete live library/DDL/repair/bulk naming acceptance remain unfinished; no live mutation or hold release is claimed.

- Delivery and next combined recovery batch (2026-10-09): PR #275 is merged at `32344cfd`; primary and integration checkouts, GitHub and the authorized mirror match. Required hosted checks are restored after the user-authorized CI waiver; local CI and commit hooks pass. The next coordinated batch combines portable parent, corrected native observer, phase-custody and factual terminal reconstruction. Peer review reproduces a late reader-pair signature refresh in the phase proposal while original input/custody bytes remain unchanged; that proposal is held pending immutable original-vector correction. Actual protected parent/daemon/NFS/terminal and full live recovery acceptance remain open. No lifecycle, library, reader database, DDL or bulk naming change ran.

- Combined parent and terminal successor checkpoint (2026-10-09): integrated the immutable original-vector phase correction, unique physical worker-bind mapping and finite critical runtime comparison. Independent terminal review repeats the prior late-helper failure and now observes refusal before ACK; the complete final observational closure follows all helpers. The combined host runner passes 277 checks with zero skips. Native/worker destination differences no longer cause a false hold; ambiguous mappings and critical process/state/mount changes still refuse. Genuine installed terminal factory, protected parent, terminal producer/scheduling, NFS and full live acceptance remain open. No live service, library, reader database or DDL mutation ran.

- Combined candidate-image proof (2026-10-09): the unchanged accepted image passes all 721 cohort checks, 277 composed recovery checks and eight runner controls with zero skips in Python 3.10.12. Fresh parity confirms all 25 installed modules; genuine installed birth/lifecycle/configured-scope factories share the original pipe, with UID 1000 and no effective capabilities. Parent observations are explicitly synthetic. The new factual terminal factory and protected/NFS/live execution remain separate unfinished gates; no live hold is released.

- Terminal factory and NFS helper checkpoint (2026-10-09): the genuine terminal provider CLI passes in the accepted image through the original pipes and canonical installed SDK/Writer types. Both fresh disposable scenarios pass: forward factual observation with every right false, and late real current-database mode change refused without ACK. Each exercises 17 challenge frames. Parent and durable declarations are explicitly synthetic; owning mutation provenance, protected process and NFS/live acceptance are unverified. The corrected NFS helper passes 31 isolated selected-image controls and independent review; these use local fixtures, not the live export. A fresh native daemon probe refuses at primary API configuration validation; the cause is under investigation without changing configuration. Terminal composer/scheduling and protected acceptance remain open.

- Native saved-configuration correction (2026-10-09): actual Mylar saves API/HTTPS booleans as `True`/`False`; the read-only probe previously required numeric spellings and refused the enabled primary API. A regression reproduces that refusal. The corrected probe accepts only finite boolean spellings, still refuses disabled API, enabled HTTPS and unknown values before transport, and preserves all original process/config/source closures. Native 39, full recovery 282 and runner eight host controls pass with zero skips; independent review repeats mixed-case and unknown-value checks. The exact corrected buffer now passes against the actual existing daemon, with fresh process/config/source binding, held startup state and census revision 13. This read-only evidence changes no live setting or data and grants no recovery authority.

- Combined corrected-helper image proof (2026-10-09): a single fresh isolated candidate-image run passes recovery 282, runner eight, both genuine terminal CLI scenarios and NFS-helper 31 local-fixture controls with zero skips. The NFS V3 source pin/copy delta independently verifies exact corrected native-probe bytes and unchanged V2 ancestor, lifecycle, kernel and test fixes. Actual live NFS support remains unproved. Scheduler V1 review reproduces a late resume-intent metadata change accepted before restart; immutable V2 now refuses it without restart and passes 84 source controls, but composed mapped observer/producer and protected runtime acceptance remain pending. No live lifecycle or media/database mutation ran.

- Delivery and primitive review checkpoint (2026-10-09): PR #277 is merged and synchronized across primary/integration, GitHub and the authorized mirror after local CI and hooks pass. The exact corrected native probe observes the existing held daemon and census revision 13 without changing settings or data. Parent scheduling V2 closes the reproduced late resume-intent mutation in 84 source controls; terminal producer composition remains held after peer review reproduces same-byte original-preimage replacement and changed output metadata accepted as fresh evidence. The correction must preserve the original execute ACK, report and sealed terminal refs. Independent disposable controls reproduce prewrite ancestor redirection and a foreign child after the final journal-root census in hardlink V1. Root independently reviews immutable V2 and repeats all 29 controls plus the three original literal failures; each now refuses, leaves protected directories untouched and preserves foreign files. This closes those source gaps; selected-runtime and actual NFS testing remain pending. The active-reader canary and stopped-reader successor are distinct missing owning gates. T020–T027 remain open; no live retirement, restart, DDL import or cleanup ran.

- Terminal provenance integration checkpoint (2026-10-09): the independently reviewed correction is integrated with the original execute ACK/report and canonical terminal reference export, finite forward/rollback observation, fresh current reader-pair custody and retained private output/phase-directory identities. Root repeats 334 recovery and eight runner controls with zero skips; both prior literal failures now refuse acknowledgement. Parent admission remains disabled. Independent archive rollback review identifies an initial durable-write lost-acknowledgement path returning the wrong error phase. The integrated correction now enters Held/uncertain while retaining both markers; root repeats both literal failures, 28 rollback, 62 adoption and 13 installation-source controls without skips. The final corrected candidate image and genuine owning runtime remain pending; the first isolated build is diagnostic evidence only. No live service, library, reader database or DDL mutation ran.

- Final selected-image verification (2026-10-09): the corrected isolated image passes one combined run of 433 unittest controls, both genuine canonical terminal CLI scenarios and 64 local-fixture NFS-helper controls, with zero skips, Python 3.10.12, UID 1000 and no effective capabilities. The private launcher now includes the required adjacent terminal helper and retains CLI errors; no production dependency or fixture type was replaced. Fresh installed-origin and source-map checks pass. Parent/execute declarations remain explicitly synthetic in the CLI harness, and NFS controls use local fixtures. Actual owning terminal provenance, approved live NFS, connected parent admission, reader indexing and library/DDL acceptance remain open. No live service or application data changed.

- Connected recovery review (2026-10-09): The active-to-stopped NFS bridge now has an isolated fixture successor that copies source inputs privately before fault injection. Parallel host controls preserve every original bundle hash and full file signature. This closes the fixture interference gap only; real export, stopped-parent admission, current reader indexing and DDL recovery remain unverified.

- Corrected recovery checkpoint (2026-10-09): independent review clears the existing-preparation successor after its final claim loop stops invoking a replaceable file-type helper. Root repeats 228 preparation/adoption/rollback controls and integrates the coherent owning installer and source closure. The NFS bridge fixture successor passes all 163 controls in the selected Python 3.10.12 image, UID 1000, with its source bundle read-only and original hashes/full signatures unchanged; this is local-fixture runtime proof only. The corrected full-policy verifier and dedicated archive provider remain under independent review. Actual protected NFS, parent admission, reader indexing, repair/import and current-library acceptance remain open.

- Connected archive image/provider checkpoint (2026-10-09): the corrected native candidate passes 241 selected controls with zero skips, fresh installed-source/type checks and the unchanged disabled parent admission. Independent review clears the dedicated archive provider after original preparation metadata and completed-directory identities are carried through verification and final acknowledgement. Root integrates that provider, its corrected verifier fixtures and explicit source closure; all 374 recovery controls pass without skips and preserve original public sources. A test-only bundled-dependency name now avoids collisions with host-installed libraries. Native/reader lifecycle composition and original directory-custody transfer, real NFS and protected/library/DDL acceptance remain unverified.

- Portable NFS integration (2026-10-09): root verifies frozen source hashes and preimages, independently reviews the complete original-mount projection correction and integrates the portable active/stopped bridge with explicit source closure. Config-file and Writer shadow mounts and ambiguous mappings refuse. The combined private-copy runner now expects 460 controls; actual NFS, archive-only parent composition and live acceptance remain pending. No service restart or data change ran.

- Combined recovery verification (2026-10-09): all 460 integrated recovery and eight runner controls pass in Python 3.10.12 using the exact corrected candidate, with zero skips, preserved original public source and no live mounts. The independently reviewed neutral archive backup producer successor now retains every original source/backup/restore file tuple before callbacks and binds its exclusive output directory before the first write. Root repeats its 22 controls and integrates it; all 482 combined host controls pass without skips. Parent custody/scheduling, actual NFS and live library/DDL acceptance remain open.

- Final combined source checkpoint (2026-10-09): the exact corrected candidate passes all 482 recovery and eight runner controls with zero skips in Python 3.10.12, using only frozen public source and private disposable fixtures. Full staged local CI and history secret scanning pass; no secrets or skipped files are found in the publication-boundary scan. Public contract commands now name the actual combined runner. Actual installed archive parent, protected NFS, reader indexing and complete live acceptance remain unfinished. No live hold is released.

- Typed archive custody integration (2026-10-09): PR #279 is merged and local/GitHub/mirror main match. Independent review clears the same-custody binding, original metadata/directory transfer, provider insertion and distinct archive-only NFS readiness. Root integrates the four sources, coherent lifecycle/admission/factory pins and exact known-predecessor installer correction. Fourteen binding and 493 combined host controls pass without skips. A fresh candidate build and genuine installed-pipe harness are pending; connected parent scheduling, real NFS and live recovery remain open.

- Installed lifecycle batch (2026-10-09): the fresh candidate image passes 377 selected build controls without skips and fresh installed module/origin/type parity. Independent integration review passes 493 recovery, 18 installer and 14 binding controls without skips; all 157 recovery source pins and the canonical cohort match. The genuine installed binding harness remains pending after test-input corrections; synthetic parent declarations grant no live authority. The ordinary import audit reproduces single/pack completion diagnostics from preexisting files without a delivery acknowledgement and identifies issue-ID-only handoff retirement. Durable owning import acknowledgement and its queue/workflow consumers remain required; NZB handoff exclusion from download retry already passes. Task order and completion gates are unchanged. No live service, library, reader database or DDL mutation ran.

- Installed binding verification (2026-10-09): the corrected genuine selected-image harness passes all three scenarios at UID 1000 with no live mounts. Canonical installed birth, custody, scope, Controller/Writer and preparation factories retain the same original pipe through 24 frames. Late metadata and completed-stage changes both Hold and invalidate custody. The SDK map uses 91 canonical filenames and fresh actual installed hashes; workflow fixtures are created by the installed owning Store. The same candidate passes all 726 cohort controls, 493 recovery controls and eight runner controls without skips in Mylar Python 3.10.12. Full staged local CI passes, including repository validation and history/staged secret checks. Parent/runtime/backup observations remain synthetic; connected parent, actual NFS, import acknowledgement and live acceptance are still required. No live hold release or data mutation ran.

- Parent composition review (2026-10-09): the unintegrated proposal passes 149 isolated controls in place, relocated and in the actual selected Mylar Python runtime without skips. Independent literal checks nevertheless reproduce late request-ancestor capture and read-only interpreter/runtime overlays. The proposal is Held and its immutable correction is under implementation; no parent pin, lifecycle action or completion checkbox is enabled by those tests. The full parent SDK closure also needs fresh protected-daemon guard source pins beyond the smaller binding-harness closure.

- Parent correction checkpoint (2026-10-09): independent review clears the immutable correction after all original ancestor and interpreter/tool/loader overlay probes Hold. All 154 controls pass in place, relocated and in the actual selected Mylar Python runtime without skips. Integration still requires coherent active backup-helper, rows, producer, adapter and admission source pins; existing negative-five compatibility must remain intact. Operational parent, NFS, backup, import and live acceptance stay open.

- Coherent parent integration (2026-10-09): root integrates the independently cleared archive parent correction and the frozen active backup-helper dependency successor. The explicit recovery closure has 167 files and 539 distinct controls; all 729 cohort and 539 recovery controls pass with zero skips, and repository validation passes. Independent dependency review clears the exact source graph. All 539 recovery and eight runner controls also pass in the selected Mylar Python 3.10.12 image, without skips; frozen source hashes remain unchanged. The first isolated runner check caught its historical 493-count expectation; the corrected count-only test passes in a new immutable runtime bundle. Historical accepted admission bytes remain unchanged, with the exact current predecessor added to the installer. Parent pins remain disabled; installed-image, protected backup/NFS, import and live acceptance remain open. No live data or lifecycle change ran.

- Ordinary import review (2026-10-09): independent review holds the unintegrated acknowledgement batch after reproducing acceptance of an identical replacement destination inode and late history/catalog-companion changes. Its 207 passing host checks and 58 reported runtime/tool skips do not accept these gaps. A new immutable correction must retain actual exclusive-copy file-descriptor identity and original passive history/companion vectors through final closure. Archive terminal history remains separately required; no import or live acceptance is claimed.

- Reviewed recovery image checkpoint (2026-10-09): the 326-input frozen public candidate passes 380 selected build controls without skips. Fresh final-image verification at UID 1000 confirms canonical installed origins, exact cohort hashes and shared SDK types; admission remains disabled. Actual image identity is `sha256:6eeda350d4515e7c6b857e5e1d61489d421358791ab57f1844f3748e24d73be2`. Full staged local CI passes; the subsequent public-contract command correction passes focused hooks, with unrelated no-file hook skips reported separately. Protected parent, NFS, import and live acceptance remain open.

- Import acknowledgement correction (2026-10-09): independent review clears the immutable successor after the original destination-inode, history and catalog-companion regressions refuse. Root verifies all 163 source/preimage/evidence pins and integrates the 22-path batch, with the current SDK source closure updated coherently. The proposal passes 212 host controls and reports 58 native/tool skips; actual installed importer execution and combined native/worker image verification remain required. Archive terminal history and live recovery are still open.

- Pack sidecar preservation (2026-10-09): root integrates the independently cleared phase correction. Sidecars count toward completion and enter cleanup only after verified preservation; a review sidecar cannot complete a pack or lose its source. Both predecessor failures reproduce in disposable fixtures, and corrupted retention bytes Hold without deletion. The combined proposal passes 214 host controls with 58 explicit native/tool skips; actual image and importer gates remain required.

- Integrated acknowledgement/pack verification (2026-10-09): root repeats 42 native history and pack-record controls successfully on the combined checkout. A separate migration run fails three controls because its retained public PostProcessor fixture is missing from the repository/image closure; this test-packaging gap is being corrected before the new image build. Worker and installed importer checks remain distinct gates. No live recovery or hold release ran.

- Combined acknowledgement/preview integration (2026-10-09): independent V4 review clears all nine source pins and repeats three prior fault refusals, seven focused native controls and 106 worker controls without skips. Root integrates the exact native derivation/approved-map join, manually coalesces image and module documentation, refreshes all 66 canonical SDK closure pins and verifies preview capability availability depends on its installed route and executor. The portable migration fixture gap is corrected and all five migration controls pass; integrated naming controls pass 106. A frozen 496-input native image build now runs full installed checks. Selected image, terminal-history integration, delivery and coordinated live acceptance remain open; no live service or library change ran.

- Selected-image integration follow-up (2026-10-09): the new isolated build applies the full patch set and passes the 20 ordinary-history and five portable migration controls, then fails an older release-naming pack-confirmation test whose fixture has no native catalog/owning import acknowledgement. The failure is retained and the image is not accepted. Review now also checks acknowledgement continuity across legitimate owned renames and passive status failures. Repository validation passes using the system Python; the first environment-selected Python lacked PyYAML. Live services remain unchanged.

- Import/normalization continuity finding (2026-10-09): independent review confirms the new durable import acknowledgement binds its original path and complete archive signature, while legitimate owned rename and metadata publication change those facts. A separate successor correction is being implemented through actual pre-transition custody and owning terminal closure, preserving the original acknowledgement unchanged. Historical tokenless packs remain review; missing catalog observations must retain members as review. The isolated importer harness passes owning registration after disposable fixture correction, then exposes missing daemon-initialized encoding in its fixture; its next corrected run is pending authentication. No installed-import success or live pack recovery is claimed.

- Actual single-DDL importer component proof (2026-10-09): the corrected disposable harness executes the installed PostProcessor.Process_next with real Config, 25 schema statements, Controller/Writer, placement helpers, catalog upsert, acknowledgement and cleanup. Success completes with no guard review and removes the source only after acknowledgement. A separate process injects a genuine postcommit lost response; the original source remains and passive exact acknowledgement confirms it without replay. Both cases pass in the selected diagnostic image at UID 1000 without live mounts. This accepts the regular single-DDL component only; full image gate, complete discovery/annual/pack/coordinated paths, normalization continuity and live recovery remain open.

- Combined worker image verification (2026-10-09): the frozen acknowledgement/preview worker passes all 562 controls with zero skips using the actual archive/PDF runtime. A fresh final-image check verifies all 30 installed production file hashes and safe modes, including ordinary_import_ack, at UID 1000 with no live mounts, network or effective capabilities. This image does not yet include the pending native acknowledgement-continuity or archive terminal-history successor; full integrated delivery and live acceptance remain open.

- Selected native source-control repeat (2026-10-09): all 50 publication-native controls pass without skips in the actual Mylar Python runtime using its installed source and archive backend, at UID 1000 in a read-only isolated container. This covers the native suite omitted on the host; it does not override the failed full-image legacy pack gate or accept future continuity/history changes.

- Parallel immutable review checkpoint (2026-10-09): Root verifies all 289 continuity and 531 archive-history source/evidence pins and every tracked preimage. The proposals are independently cross-reviewed in parallel before shared installer/cohort/provider coalescence. Existing actual single-DDL importer and worker proofs remain component-only and do not accept these successors or live recovery. No live lifecycle or media change ran.

- Reproduced review corrections (2026-10-09): Root repeats all 68 frozen continuity source controls without skips after supplying the actual archive-tool root; an earlier bare host invocation reports 14 tool skips and is retained separately. Independent review nevertheless holds three reproduced original-FD/pending/late-closure faults. Root prepares an isolated installed continuity harness; no new source or image acceptance is asserted. Live services and library remain unchanged.

- Reviewed import continuity integration (2026-10-09): root independently repeats the original pending-FD, late-pending and completion-inode fault wrappers against the immutable successor; all fire and Hold, preserving the original acknowledgement. All 289 source/evidence pins and seven preimages match. Root integrates the seven-path successor, coalesces installer/image checks and documents private-journal backup prerequisites; all 67 canonical SDK pins match. Frozen and integrated repeats each pass 74 controls with zero skips, and fatal lint/Python 3.10 grammar pass. The first incorrectly scoped root integration command fails its preimage guard before writes; its predecessor test failure is retained separately. Full image, installed continuity/discovery, historical missing-ACK and live acceptance remain open.

## Convergence assessment and evidence

The focused recovery assessment identifies remaining build and runtime work in the existing owning import/pack boundaries. T028 and T029 are placed above with their prerequisites, preserving previous IDs and completion history. T028 source review precedes the relevant T026 image gate and T017 live reconciliation. T029 verifies the current combined cohort before T017–T019 acceptance or cleanup; historical component proofs cannot substitute for it.


- Combined source verification checkpoint (2026-10-09): Root integrates the independently cleared continuity V2 and repeats all three original inode/mode faults; each refuses, with original completion rows unchanged. Both frozen and integrated 74-control runs pass without skips. The independently cleared archive-history V2 carries original history before callbacks and closes every queue branch after its final observer; both original review faults refuse. Root integrates the source and installer successors with 68 SDK entries and 168 explicit recovery files. All 539 recovery controls pass without skips, and 25 archive history/installer host controls pass with one explicit installed-image skip. The affected batch also passes 20 rename, 32 maintenance, seven supplement, 35 combined-publication, 12 queue and 21 reader-proof installer controls. Legacy pack/tagging fixture module closure is being corrected; the NZB check rerun with the valid host context passes six controls and reports two installed-source skips. Current full-image, original namespace, protected backup, historical completed-download recovery and live acceptance remain open. No service, media, queue or database mutation ran.

- Combined candidate checkpoint (2026-10-09): Independent dependency review clears all 68 current SDK, 25 prospective module and 168 recovery source pins, with activation pins still unset. Root integrates the narrowly scoped two-fixture correction after verifying 291 frozen pins and exact preimages. Original and relocated full fixture runs each pass 134 controls with three existing image-only skips; root independently repeats all 22 pack and 13 backend controls with no skips. Corrected repository validation passes. A new 501-input native candidate and matching 56-input worker candidate are building in isolation; neither is accepted or deployed. A separately labelled diagnostic image will exercise the four independently reviewed genuine importer cases while the complete image gate runs. Annual/coordinated-pack execution and reviewed retained-delivery recovery are being developed in parallel. No live lifecycle, library, database or DDL mutation ran.

- Installed runtime checkpoint (2026-10-09): The new matching worker image passes all 567 controls without skips; final read-only UID1000 parity verifies all 30 production sources, including the archive factual-history consumer. In the separately labelled diagnostic native image, genuine ordinary import success, real lost postcommit response and actual import-to-owned-rename-to-preserved-metadata continuity pass against a fresh 97-file actual SDK map. The original acknowledgement and preservation pair remain unchanged. Complete discovery still fails a terminal-outcome assertion after its acquisition fixture was corrected; retain that failure and diagnose the actual queue result rather than relax the assertion. The installed literal nzblog schema is independently observed. The full native image gate continues; annual/coordinated-pack harness and retained historical-delivery acceptance are parallel private proposals. No live deployment or completion acceptance is claimed.

- Container-filesystem verification gap (2026-10-09): The complete native candidate gate fails the existing terminal-observation producer controls at `terminal-output-created-census`, after the owning source suites pass. A separate disposable overlay-filesystem probe reproduces stale results from reusing an enumerated directory FD after creating a file; rewinding the same original FD returns the correct census. The source correction must retain all original directory/file/ancestor and final raw closure checks, coherently update active dependency pins and pass the existing controls in the actual filesystem. The failed full candidate is not accepted or deployed. Discovery terminal diagnostics and annual/coordinated importer checks remain separate genuine-runtime work; no assertion or acceptance requirement is waived.

- Verified filesystem correction (2026-10-09): Root integrates the independently cleared same-directory-FD rewind with its coherent six-path active dependency successor. The existing 19 terminal-producer controls now pass without skips in the actual selected Python 3.10 overlay filesystem, retaining original source vectors. The original failed complete image and literal stale-directory probe remain retained evidence. A new 501-input complete native candidate is building with the correction. Genuine installed annual processing also passes with exact release identity; completed-discovery and coordinated-pack harness corrections remain under review. Independent review holds the disabled retained-delivery V2 proposal after its outer passive status fails to carry the target’s original vectors past the final helper; a new correction is required. That proposal is not installed or activated. Live services and data remain unchanged.

- Retained-delivery review checkpoint (2026-10-09): The disabled V4 proposal passes 40 source controls both in place and after relocation. It rejects all four retained original fault probes: changed private-parent mode, target mutation after passive verification, forged completion JSON without an owning catalog event, and pending Writer state introduced after the last helper. Its independent original event is currently process-local: an actual forked process refuses, while fresh genuine Controller/Writer objects in the producer process pass. Independent review is pending; the source is not integrated or installed. Durable cross-process provenance, authenticated API/pack consumers and installed acceptance remain unmet prerequisites. Complete native image verification continues in parallel. No live lifecycle, queue, library or database changes ran.

- Retained-delivery source correction integrated (2026-10-09): Independent V4 review clears the disabled same-process kernel after repeating all four original fault probes, fresh genuine Controller/Writer objects and an actual fork refusal. Root verifies all 534 frozen pins and exact preimages, integrates the four owning paths, and repeats all 40 controls with the public archive backend without skips. An initial root invocation used an incorrect tool path and skipped all 40; retain that log separately. The ongoing native image contains the prior source snapshot and does not accept this new kernel. Installer/cohort closure, genuine endpoint and worker/pack consumers, durable cross-process provenance and live recovery remain open; T028 stays unchecked and default-disabled. The existing full native snapshot passes its 539 recovery controls and continues remaining tests.

- Disabled retained-delivery packaging (2026-10-09): The owning installer now copies the new helper, the existing SDK closure has 69 exact tracked provenance entries, and image checks require canonical origin, exact helper bytes, disabled state and native/history callable seams. The complete image gate includes all 40 owning source controls; those host-style fixtures do not establish canonical installed semantic acceptance. All affected files parse as Python 3.10 and all 69 current source rows match. Parent activation pins and prospective cohort remain unchanged. Independent packaging review and a fresh matching image are pending; no endpoint or worker consumer is activated.

- Complete native snapshot accepted in isolation (2026-10-09): The 501-input native V2 build passes its complete reliability gate without reported skips, including all 539 recovery controls. Fresh final-image checks at UID1000 with a read-only filesystem and no live mounts verify exact bytes for all 93 production cohort sources and canonical SDK/type identities. The first parity harness compared the SDK list with a tuple and failed after the 93 byte checks; its corrected successor passes and both logs are retained. This accepts the prior source snapshot only: the newly integrated disabled T028 V4 and 69-entry SDK closure require their own matching image and genuine semantic checks. No live deployment, pack recovery or feature completion is claimed.

- Matching disabled-kernel candidate (2026-10-09): Independent packaging review passes 109 focused source controls without skips and confirms the new owning suite is submitted once and every result is checked. The actual installed cohort verifier now also checks exact helper bytes, disabled state and canonical module identity. Repository validation passes with the system Python. Root freezes 503 public inputs and starts the matching native V3 full image build. No live mounts or service changes are involved. Genuine retained-target installed semantics, ordinary seven-case handoff completion, durable cross-process provenance and protected live acceptance remain outstanding.

- Matching image and genuine-runtime checkpoint (2026-10-09): The 503-input native V3 candidate passes its complete image gate without reported failures or skips, including the disabled retained-target module and its 40 owning source controls. Root observes the actual finished image identity. Independently reviewed seven-case ordinary and five-case retained-target harnesses are queued for authenticated isolated execution; no genuine result is claimed while pkexec waits. Independent review holds the next finalizer proposal: its first source-reading callback can change a shared source parent or the not-yet-captured second module leaf and have those new facts adopted. Both exact fault probes are retained and a two-pass original-source correction is in progress privately. No finalizer source is integrated, no endpoint/consumer is enabled, and live state remains unchanged.

- Delivery audit and continuity follow-up (2026-10-09): The staged publication batch passes local CI, including staged secret scanning and 579-commit history scanning. The first CI run rejects trailing whitespace in the exact imported Mylar fixture; a file-specific Git attribute now preserves that fixture unchanged, following the existing imported-patch convention. The publication scan reports no secrets, no skipped files and no privacy findings in the staged paths. Independent delivery review identifies an outstanding worker integration gap: native import confirmation consumes owned rename/metadata continuity, but the worker still requires the original path/signature. A separate read-only worker continuity successor is being developed; original ACKs and control evidence must remain unchanged. The finalizer V2 correction is frozen privately for independent review and remains excluded from delivery. Genuine installed runtime tests still wait for authentication; no live state changes or feature completion are claimed.

- Retained finalizer integration and review follow-up (2026-10-09): Two independent reviews clear the corrected original-source capture for the disabled same-process retained-pack finalizer. Root verifies all 520 frozen file pins and exact preimages, integrates its five owning paths and installer checks, and repeats all 38 finalizer controls without skips. The SDK closure now has 70 exact source entries; the new helper remains disabled and requires a fresh matching image. The previously accepted 69-entry image and its pending isolated checks remain distinct evidence. Delivery review reproduces a separate late catalog-WAL gap after the archive preparation route’s last history callback; its full-original-vector correction is under parallel implementation and review. Root approves the bounded authenticated worker continuity contract for private implementation, with original ACK/attempt bytes, dispatch-return custody, unlocked transport and locked consumption; saved JSON and missing witnesses must hold. Cleanup and restart provenance remain unaccepted. Task checkboxes remain open; no live queue, database, media or service changes ran.

- Expanded source validation (2026-10-09): Root repeats 38 finalizer, 40 retained-target, 21 reader-proof installation and two cohort controls successfully without skips; repository validation passes for the expanded working tree. The separate installed archive probe has one expected host-only skip and does not prove installed acceptance. The two-path route correction is frozen with 129 portable and 129 relocated passing controls, pending independent review before integration. A genuine installed finalizer harness is being prepared against the next 101-source image map. The earlier staged CI result applies to its earlier snapshot; the final coalesced batch still needs fresh delivery checks and installed runtime verification. No live handling or completion checkbox changes are claimed.

- Archive route correction integrated (2026-10-09): Independent review clears the frozen two-path correction after the unchanged original late-WAL probe and seven additional late mutations all hold; 134 review controls pass without skips. Root verifies the frozen manifest and exact source preimage, integrates the route and 18-control owning suite, updates its canonical SDK source pin and adds the actual image-gate suite. All 18 integrated route controls pass without skips. The route preserves detached original preparation/history vectors and closes namespace, claim, absence, file/node and Writer facts after the last helper. Retained reader-recovery fixture pins remain unchanged; no old proof is relabeled. A matching final image and genuine runtime remain required. The worker continuity bridge is still under implementation and separate review; all live holds remain unchanged.

- Coalesced source checkpoint (2026-10-09): Repository validation passes after finalizer and route integration. Root has repeated 144 focused controls without skips; the distinct installed-only archive probe remains skipped on the host. The expanded publication scan has no secrets or changed-path privacy findings. Worker continuity implementation and its independent review, matching installed-image acceptance, protected live recovery and final delivery remain open.

- Genuine-runtime results and purpose correction (2026-10-09): The authenticated isolated runs complete. Ordinary success, lost response, owned continuity and annual cases pass; completed discovery and both coordinated-pack cases fail their full-Process queue predicates and remain under exact source/trace diagnosis. All five retained-target cases stop at held-ordinary-writer before acceptance. Source inspection identifies a real composition mismatch: native_writers.operation admits pending/recovery purpose flags, whereas the retained owning kernel requires all flags false. A bounded opt-in ordinary-purpose operation and corrected wrapper composition are being developed privately, preserving default operation/recovery behavior and the kernel checks. No fixture bypass is accepted as native-wrapper proof. The genuine finalizer harness is frozen privately but needs matching corrected sources and installed execution. User authentication is requested for read-only public API extraction from the accepted candidate image to verify the bridge against exact installed source. No live service, queue, database or media changes ran.

- Runtime diagnostic and harness review (2026-10-09): The private diagnostic-only importer successor preserves every original fixture and acceptance predicate while adding bounded row logs and actual duplicate-check/Process_next flow observations. Root audits its frozen delta and queues the three failed cases against the original accepted image with exact source-map cardinality. No duplicate success or historical ACK is invented. Independent finalizer harness review clears its exact 101-key/95-production-pin contract and 20 host plus 20 relocated controls; the known native-wrapper purpose mismatch still blocks positive runtime acceptance, so an explicit corrected source-pin successor and all eight installed cases remain required. GitHub currently reports no open PRs. Shared worker/native-purpose corrections continue privately; no source or operational acceptance is implied by the harness review.

- Ordinary-purpose wrapper correction integrated (2026-10-09): Independent review clears the opt-in strict native operation after 94 public source controls and 18 additional purpose/cleanup checks pass without skips. Root verifies frozen pins and exact preimages, integrates native_writers/publication_native plus the 16-control owning suite, repins both canonical source entries and submits the suite to the image gate. All 16 integrated controls pass without skips. Only the four disabled retained-target/finalizer wrappers opt in; the strict owning kernel and default legacy/recovery behavior remain unchanged. A new matching image and genuine five/eight-case runtime results are required. Exact actual importer/helper bytes are independently reproduced from retained public sources and frozen accepted patch functions, matching the previously captured runtime source-map hashes; these are reproduction evidence rather than privileged extraction. They identify duplicate preflight in the old existing-target fixture, so a genuine first-import fixture is being designed before admission. No historical ordinary ACK or live cleanup is manufactured.

- Worker continuity correction checkpoint (2026-10-09): Independent review holds the uninstalled V1 bridge because the actual normalizer marker blocks its lifecycle, mapped native missing-file claims are omitted, and its response witness retains reconstructed JSON instead of the original HTTP envelope. The frozen V2 correction contract is reviewed against the owning cycle: collect selectors without authority during the marked cycle, then obtain fresh strict admission after all existing dispatch, refresh, reader and naming work, retain the exact authenticated response bytes, and consume immediately under the writer with complete mapped claim closure. Implementation and independent review remain in progress; no marker exemption, old-proof rewrite, activation or live completion is claimed. The authenticated installed diagnostic confirms all three retained duplicate fixtures enter actual dont_dupe paths without Process_next; its strict terminal assertions remain unchanged. A separate first-import fixture is being constructed before registration with a genuine incomplete target and distinct registered seed.

- Combined delivery checks (2026-10-09): The explicit 74-path staged source passes full local CI and repository validation. Publication scanning covers 2,819 files with no secrets or skips and no privacy indicators in changed paths. The current 507-input isolated native image is building against the freshly verified parent; matching retained/finalizer harnesses keep all 13 semantic scenarios unchanged and exactly match the current 95 production-source pins. Independent review repeats 33 portable harness controls without skips. These checks do not establish installed runtime acceptance, live recovery or completion of any remaining rollout task.

- Genuine first-import runtime checkpoint (2026-10-09): The corrected disposable fixture uses the installed ComicSize TEXT schema, an initially Snatched/unplaced target and a distinct registered Downloaded seed. Three original-image scenarios pass actual placement, catalog update and durable acknowledgement: regular success, lost response and owned naming/metadata continuity. Discovery, annual and both coordinated-pack scenarios remain unaccepted. The annual reaches a guarded rescan refusal; the full queue cases need exact owner-row and outer-log diagnostics before any protocol correction. The prior duplicate and schema-failure runs remain retained. A four-case diagnostic successor preserves fixture and success checks. These results cover the original accepted image only, not the new source/image or live recovery; T028/T029 remain open.

- Independent bridge review (2026-10-09): The frozen V2 bridge remains uninstalled after a literal late-change probe reproduces an accepted mapped native journal mode change: the receiver compares the original hash/signature, then captures a fresh leaf vector. Its correction must map and retain all native original file vectors before replaceable receiver callbacks and conflict-join later captures rather than adopt refreshed evidence. The actual cycle, mapped claims and raw envelope corrections do not waive this additional finding. The current 507-input native image separately passes 2,624 printed tests across 126 suites plus 539 recovery controls, with no reported skips or failures; actual image identity and 13 installed semantic cases still require authentication/execution. No live success or task completion is claimed.

- Coalesced implementation checkpoint (2026-10-09): The four actual diagnostic cases confirm that full Process emits an owned stop followed by an outer stop copying the same successful log; lost-response review can propagate before the outer stop. A source-backed disposable-fixture successor is being prepared without inventing acknowledgements. The finite retained finalization/status endpoint is integrated after independent source review; shared source gates, worker receipt consumption, restart provenance and live acceptance remain open. The preceding native image passes its complete isolated build gate and has a verified Docker identity; its 13 retained/finalizer cases are queued separately. No live recovery or completion checkbox changes are claimed.

- Shared source integration (2026-10-09): Independent review clears worker bridge V3 after 203 source controls pass without skips and both unchanged late-change probes refuse. Root verifies all frozen inputs and integrates the native/worker owning paths, coalescing the finite retained endpoint classifier with the separate read-only observation command. The retained endpoint also passes 120 repeated local controls without skips. Shared installers, canonical source pins and combined image gates are being reconciled; these source checks do not prove installed transport, durable pack completion, source cleanup or live recovery.

- Combined source verification (2026-10-09): Shared installers and image gates now match all 72 SDK and 25 prospective reader source pins. Root and independent review each pass 162 native plus 40 worker controls without skips; repository validation passes. Activation remains disabled. The publication audit covers 2,842 files with no secrets, no skips and no privacy findings in the changed paths. Local CI initially flags three public API source SHA256 digests as credential patterns; independent review verifies the exact file digests, and three line-specific fingerprint exceptions preserve the checked source bytes. The corrected full local CI run passes, including staged secret scanning. Matching installed images, authenticated transport, durable retained-pack recovery, reader acceptance and live rollout remain open.

- Reviewed delivery and installed diagnostics (2026-10-09): PR #282 merges the reviewed default-disabled source batch; GitHub, local main and the authorized Gitea mirror match `149550ebd18b73c23eb626d431bc60bc3319e25e`. Both non-empty commit-range secret scans and normal commit hooks pass. The user-authorized hosted CI waiver is limited to that merge; all seven original strict check requirements are restored and read back. Six of seven genuine first-import cases pass in the older accepted 99-source image. The pack lost-response fault does fire after its acknowledgement, but the harness retention check follows a staging path that the actual processor renamed; an original-attempt-path successor is under review. The preceding 101-source retained/finalizer runner stops before its cases at an upstream source-permission check; actual image attributes are pending inspection. Neither image proves the current 103-source cohort or live recovery. Retained-pack consumer completion and original-child archive terminal augmentation continue privately in separate units. No activation or live mutation ran, and remaining task checkboxes stay open.

- Next pack-consumer review (2026-10-09): The private V1 consumer passes 88 owning controls without skips but remains unintegrated after independent review reproduces two defects: directory enumeration can precede capture of known control-file originals, and hashing every library file imposes a 32 GiB collection limit. Its correction must capture known controls before enumeration, retain complete original file/node/namespace/claim evidence and hash only affected files. The frozen failure evidence is preserved. Separate original-child archive terminal augmentation is under implementation; it supplies no cross-container producer authority. No current-image or live acceptance is claimed.

- Original-child review checkpoint (2026-10-09): The unintegrated two-module archive terminal proposal passes 199 host and 199 relocated controls without skips, but independent literal probes still return stale evidence when the last directory enumeration changes the original writer purpose, removes the actual export registry entry or swaps the original pipe descriptors. Its successor must recheck the captured logical identities and pipe custody after all physical checks, preserving original vectors. The failed proposal and probes remain retained; no installed or live acceptance is claimed. The previous three isolated authentication requests remained waiting and were stopped. A fresh read-only retained/import test request is pending; no service or library changes ran.

- CI correction checkpoint (2026-10-09): Hosted main checks expose imported upstream snapshots selected by Source lint, excluded build-only worker fixtures and archive-dependent controls run against the pristine upstream image. The candidate preserves exact snapshot bytes with three narrow lint exclusions, includes the public fixtures only in the checked build stage and selects explicit upstream/custom API suite families. The pinned shared selector checks all 593 maintained Python files successfully. With the public archive backend, all 162 API/retained controls pass without skips; the upstream-only 42-case selection reports four passing controls and 38 backend-dependent skips. Corrected local CI and repository validation pass; the earlier missing-PyYAML invocation is retained. Independent review and hosted build verification are pending. No current image, library or live recovery is accepted by these checks.


- Pack and archive verification checkpoint (2026-10-09): The private pack-consumer correction passes 95 controls and independent review repeats its original control-capture and large-library faults successfully. A composed real-factory HTTP test now reaches native finalization and exposes a JSON list/tuple mismatch in worker consumption; the successor remains under implementation. The archive original-custody successor passes 208 controls in root's independent run, and all five unchanged late writer, registry, pipe and logical-core probes refuse stale evidence. Larger same-child forward/rollback history and an observer using existing read-only worker mounts remain private; no current-image or live acceptance is inferred. Activation stays disabled and task checkboxes remain unchanged.

- CI follow-on (2026-10-09): PR #283 is merged, and GitHub, local main and the authorized mirror match `9e42db7c669f9db7507395dbd3439dcc05c3d46a`. Source lint and the worker build pass. Hosted upstream verification exposes two additional owning suites requiring the archive backend. The candidate explicitly reports 38 finalizer skips and three native-producer skips in that compatibility lane, while retaining its 13 backend-independent native-purpose controls. With the actual public backend, all 54 owning controls pass without skips. The custom image retains its mandatory backend and full-family checks. The official Moby matcher reveals that the previous directory exceptions admitted arbitrary configuration descendants. Explicit subtree exclusions and public leaf exceptions now include all 470 required source/fixture entries and exclude all seven private-configuration sentinel paths. Root and independent review each pass the complete 162 copied-source API controls without skips. Full local CI, repository validation and the 2,843-file publication audit pass; the audit reports no secrets or skipped files and preserves 180 unrelated baseline privacy indicators. This is source acceptance; actual image verification remains pending. No service, library or queue recovery ran.

- Composed review findings (2026-10-09): The genuine pack workflow now completes HTTP finalization and worker receipt consumption in private controls, but a hostile escaped snapshot can exceed the actual response limit after commit. A successor checks that serializer before finalization SQL and preserves both original size limits; matching-image proof remains pending. Root independently runs all 26 worker-observer controls, then reproduces a late claim callback that changes an already-checked archive while a final reply still escapes. That uninstalled proposal is held for a complete final physical closure, with its original probe retained. These findings keep combined runtime, recovery and live acceptance tasks open.

- Larger batch integration checkpoint (2026-10-09): PR #284 is merged at `50f56738bae5decdbd348c68d72efd502894bb4a`; GitHub, local main and the authorized mirror match. The original seven strict required checks and all other protection settings were restored and independently read back after the approved hosted-CI waiver. Root integrates only the independently reviewed disabled pack V2 and archive-originals V3 bases, plus worker-observer V2. The observer passes 30 root and 30 relocated controls without skips, and the unchanged original late-claim fault now holds with no final release acknowledgement. Pack composed V3 passes 108 controls but independent real-HTTP review reproduces a final filesystem callback that changes the original mapping while receipt consumption still succeeds; that proposal remains held and excluded, with the literal failure preserved. Shared finite fixture-copy, build-context and runtime registrations are being coalesced; combined source and matching-image checks remain pending. Same-child parent/history composition continues privately. No live state changed, activation remains disabled and completion boxes remain open.

- Upstream image gate follow-up (2026-10-09): Hosted validation for PR #284 passes the retained backend eligibility and prospective reader suites, then fails the unchanged exact-origin check because its synthetic upstream build leaves `config.py` and `Mylar.py` group-writable. The production Dockerfile already closes those modes to 0644 and 0755. The next combined candidate applies those same two mode changes in the CI-only build instead of weakening the verifier; the existing custom-image workflow controls pass. Actual hosted rerun remains pending.

- Integrated archive source checks (2026-10-09): After root coalesces the accepted lifecycle source pins, 275 copied-source archive, lifecycle, preparation, admission and installer controls pass without skips. The first run rejects a stale source assertion; the second rejects a historical fixture used as the immediate predecessor. Their logs are retained. A new exact immediate-predecessor fixture preserves the prior admission bytes, while the historical fixtures and their upgrade coverage remain intact. These checks cover the integrated disabled base only; same-child parent composition and installed/live acceptance remain outstanding.

- Composed pack source acceptance (2026-10-09): Independent review clears the V4.1 correction after 111 host and 111 relocated controls, including the unchanged final-mapping fault. Root integrates its finite 35-file build-only fixture graph and exact source pins. Root and peer each pass all 184 copied-source API controls without skips; upstream eligibility explicitly skips the 22 new backend-dependent cases. The final-mapping fault now refuses acknowledgement, but its post-CAS receipt and intent remain retained in a consuming state. Recovery of that uncertainty, operational initialization and lost-reply status handling remain under implementation. The archive same-child batch is frozen for independent review after 493 host and 493 relocated controls; it is not yet integrated. No live changes ran and completion boxes remain open.

- Reviewed source delivery (2026-10-09): PR #285 merges the disabled pack/archive source batch at `9d0384e7dcd3db313ef802a343ee87fe02ac45ae`. Full local CI, normal hooks and both non-empty commit-range secret scans pass; the publication audit scans 2,887 files with zero secrets or skips. All seven strict hosted check requirements and every other branch-protection field are restored and read back exactly after the authorized waiting waiver. Independent review of the separate same-child proposal reproduces a final pipe-descriptor callback changing the original archive after its last file check while an acknowledgement still succeeds. That proposal remains excluded and its correction is under implementation. No deployment or live recovery is accepted; remaining task boxes stay open.

- Operational review findings (2026-10-09): The separate pack operation proposal passes 126 host and 126 relocated controls, including one finalize request followed by one same-live status request after a lost reply. Independent review reproduces an existing-journal capture defect: an admission callback can change its original attributes before initialization captures them. That proposal remains held for first-capture correction; it is not integrated or deployed. The same-child archive proposal also remains held for its final-descriptor callback correction. Original reproductions are retained and source acceptance does not complete T028/T029 or live recovery.

- Pack operation source integration (2026-10-09): Independent review clears the corrected explicit operation batch after 130 source and 130 relocated controls pass without skips and the unchanged existing-journal fault now refuses. Root verifies the complete 711-pin freeze and original V4.1 preimages, preserves the existing backend eligibility decorator, integrates ten owning paths and expands the finite build-only graph to 36 entries. One live finalize attempt may use one read-only status request after a lost reply; the original action and full consumption checks remain required. The local view reports zero imported members. Combined source checks, authenticated installed transport, outer scheduling, restart and post-CAS recovery remain open; no live changes ran.

- Combined operation and same-child source checks (2026-10-09): Independent review clears the corrected archive same-child batch after 503 source and 503 relocated controls, including the original final-descriptor fault. Root integrates twelve owning paths and updates the exact cohort, installer and finite control graphs; existing accepted installer predecessors remain intact. Independent packaging review confirms 337 API source entries and 36 build-only pack fixture entries. The combined copied reader run passes 1,288 tests without failures, and full local CI passes. The API run passes 199 tests but fails its final source-attribute closure because a concurrent root fault-test invocation changes and restores an original module mode. That run is not accepted; the unchanged sequential repeat passes all 199 controls and the final original-source closure. The additional same-child suite passes 83 controls in a private copy. Matching installed transport, worker observation, external operation routing and live recovery remain open. No live state changed and task boxes remain open.

- Reviewed combined source delivery (2026-10-09): PR #286 merges the 40-file operation and same-child batch at `faa5776b02203d16c99678483979c272aef77599`. GitHub main, local main and the authorized mirror match the reviewed tree; all seven strict check requirements and every other protection field are restored exactly. Staged local CI, normal hooks, 642-file Python correctness lint, publication audit and both non-empty commit-range secret scans pass. Runtime activation and live recovery remain open. The next parallel units implement archive worker observation and a separate retained-pack parent conversation that preserves one worker session across actual backup and isolated restore. A simple CLI cannot establish that ordering; historical standalone delivery needs a distinct retained-repeat producer. No live data changed and no acceptance box is completed by this source delivery.

- Combined worker observation checks (2026-10-10): Independent review clears the separate archive-worker observation composition and its finite build-only fixture packaging. Root passes 199 copied-source API controls with final original-source closure and 588 copied recovery controls across 31 suites without skips; additional copied native and worker checks pass 116 and 43 controls. All 174 recovery and 337 API source pins match. The first recovery run exposed a repository-relative test fixture path; the corrected run uses one exact adjacent production-source fixture and preserves the original assertions. Matching candidate images, installed transport, NFS evidence and live acceptance remain open. Activation stays disabled and remaining task boxes stay open.

- Retained operation and tooling checkpoint (2026-10-10): All 13 historical installed retained/finalizer cases pass after seven controller-fixture arguments use the production string-root type; production code and success assertions are unchanged. That historical image does not prove the current source cohort. Independent review holds the separate retained parent proposal for final source closure and premature worker exit; its successor remains private. Two reviewed development-tool dependency fixes are integrated locally, and the actual Node 26 image, post-create setup and smoke checks pass. No live service or library changes ran.

- Candidate build corrections (2026-10-10): The combined image gate refuses acceptance when all 28 composed pack tests lack the installed archive backend path. The checked Mylar stage now exports that existing fixed path. Two worker fault controls also require changing their original source attributes, so the checked worker stage runs the unchanged full suite in a private user-owned copy and keeps the installed source tree intact. The failed build is retained; the corrected two-image gate is pending. Full local CI, repository validation and 646-file Python correctness lint pass for the preceding source batch; its publication audit scans 2,897 files with zero secrets or skips. No installed or live acceptance is claimed.

- Reviewed worker source delivery (2026-10-10): PR #287 merges the 30-file default-disabled worker observation, build-fixture and development-tool correction batch at `3b084738f67e8d3cb42fb75ec7e472e9ebade9a4`. GitHub, local main and the authorized mirror match the independently reviewed tree. All seven strict hosted check requirements and every other branch-protection field are restored exactly. The corrected full local CI, normal hooks and both non-empty commit-range secret scans pass; local development dependencies are installed from the updated lockfile. The corrected isolated Mylar/worker image gates and the separate retained-parent successor review remain in progress. No live application update or recovery is accepted by this delivery.

- Retained-parent integration and runtime checks (2026-10-10): The corrected worker image passes all 679 controls and an isolated read-only parity check of 34 installed modules. Mylar passes the 199-test copied API family, then its image gate refuses two redundant direct API suites that lack their generated fixture; independent review clears routing those suites only through the mandatory copied API family. Independent review also clears the retained-pack parent after 178 source and 178 relocated controls without skips; all six original fault probes refuse. Root integrates its default-disabled parent, backup, worker control and exact handoff successor, expands the finite copied API graph to 346 entries and the build-only worker graph to 37 Python files plus its pin map. The combined API check is running; the unchanged 588-check recovery gate passes. Fresh read-only inspection confirms Mylar remains unhealthy, Komga healthy and the normal worker held. Standalone recovery is being implemented in parallel against the reviewed original-source, complete SQL and four-role backup contract. No live update, recovery, reader resume or task completion is accepted by these source checks.

- Merged source and capacity checkpoint (2026-10-10): PR #288 is merged at `adc6d17cc21c29cc7303bfea7db38fe0162fe146`; GitHub, both local checkouts and the authorized Gitea mirror match. Full local CI, normal commit hooks, the 2,907-file publication audit and both non-empty commit-range secret scans pass. All seven strict hosted check requirements and every other protection field were restored exactly. The immutable combined image build passes all 679 worker controls and exports its worker candidate, but native verification refuses an original archive readback while preparing a Boolean-sequence fault fixture; diagnosis remains open and no rollout acceptance is inferred. A read-only detached database sizing check verifies original application files unchanged and measures 34,180,534 bytes of complete typed SQL, exceeding the standalone draft's 4 MiB wire limit. Separate private agents are designing a bounded full-evidence transport; it is not accepted or installed. Root also integrates the independently reviewed two-file owning recovery-gate correction and its exact API pins: private source and relocated runs each pass 8 owning, 738 prospective and 588 recovery controls without skips. Root combined and matching-image checks remain pending. No live service or application data changed; activation and all operational completion boxes remain held.

- Hosted CI requirement resumed (2026-10-10): The user revokes the outage waiting waiver. All future merges require successful hosted CI; no branch-protection bypass is authorized. Hosted validation on merged `adc6d17` exposes read-only fixture mutation failures in the native-image and comic-maintenance jobs. The correction must preserve the original read-only mounts and every owning fault assertion while supplying private disposable fixture copies. Root's current combined source run passes all 738 prospective and 588 recovery controls without skips, and repository validation passes; these local checks do not replace hosted acceptance. Diagnosis and CI repair continue in parallel with standalone recovery implementation.

- Root combined and installed worker verification (2026-10-10): The integrated source passes 247 copied-source API controls, 738 prospective controls, 588 recovery controls and the separate 8 owning controls without skips. Full local CI and repository validation pass. A separate actual Python 3.10 diagnostic preserves all archive assertions and passes the unchanged 116-control suite plus ten Boolean-case repeats; it does not explain the prior full-build refusal or replace complete matching-image acceptance. The worker candidate passes all 679 build controls; root independently verifies all 35 installed module hashes and original file closure in an isolated read-only, network-disabled UID1000 container with zero effective capabilities and no-new-privileges. Its retained-pack activation remains disabled. The privately reviewed complete-body transport successor preserves full typed SQL and original file descriptors under fixed bounds; implementation and actual fit remain pending. Hosted CI fixture repair and full native candidate acceptance remain open, as do live recovery and operational task boxes.

- CI isolation verification (2026-10-10): The user’s restored hosted-CI requirement remains mandatory for every future merge. The reviewed launcher retains read-only source mounts and hardened container flags while running deliberate source-mutation fault tests on private copies. Root’s actual worker run passes all 679 tests without skips. The native owning gate now gives each of its unchanged 28 parallel suites a separate finite source closure; its exact API source pin is updated. Full local CI passes. Complete native-image verification, the existing 64 MiB offline tmpfs limit and final-head hosted CI remain pending; no live recovery or operational completion is claimed.

- Offline CI capacity correction (2026-10-10): The actual native gate refuses four startup fixtures because the existing 64 MiB temporary filesystem cannot satisfy the unchanged 128 MiB preservation reserve. Increase only its bounded test allocation to 256 MiB temporary storage and 512 MiB memory; retain network isolation, UID1000, read-only inputs, dropped capabilities and every preservation assertion. The full repeat and hosted successor checks remain pending.

## Retained execution context before compaction (2026-10-10)

T001 → T002–T003 → T004–T005 → T006–T012 → T013–T015 → T016 reviewed delivery → T020–T025 plus T028 reviewed owning recovery source → T026 current code/image integration → T029 combined runtime acceptance → T017 held matching-image rollout → T027 fresh stopped-reader/restore/NFS and aggregate terminal proof → finish T017 reconciliation → T018 → T019. Original task IDs are retained. Privileged access is currently authorized by the user’s latest resumption; historical pause notes below remain historical. T018 produces the shared current-library acceptance receipt also consumed by feature 024 T018; feature 022 T024 separately proves post-bulk idempotence and final data acceptance. T017–T019 remain incomplete. T009–T015 source and isolated-plan acceptance are complete. T016 independent reviews, final image gates, local CI and reviewed source delivery have passed. The primary checkout is synchronized, and matching native V6 and held worker V7 images are selected. Privileged access resumed on 2026-10-07. T017 has a verified consistent backup, exact catalog preservation, committed bootstrap authority and all 13 reviewed correction-family registrations. Remaining archive/ownership recovery and full-library acceptance are unfinished. The same native process remains held while complete current-library acceptance is pending; advisory checks alone do not release native or general worker startup. T003 startup exclusion has passed corrected-source reviews and the isolated native image gate. T004–T005 authenticated registration and isolated installation verification are complete; they do not complete startup admission or prevention.


## Authenticated image gate before complete verification — 2026-10-10

Snapshot from PR #291 commit `42ebee00d819418074108bde94bfe2b6b17cc3a6`, retained verbatim. The current task notes supersede its pending-build status.

- T026/T029: Verify the integrated standalone installer, exact source manifests and connected runtime gates. Independent source review and the earlier root API, standalone and copy checks pass; the copy regressions are selected by the mandatory image gate. The reviewed canonical launcher, action and parent are integrated with matching installer and SDK/source pins; all 176 combined standalone checks pass. All 247 combined API checks also pass. The reviewed packaging is integrated, and its selected-image Python 3.10/OpenSSL run passed six copy and all 121 recovery controls with zero skips after the final preflight correction. Full local CI and secret scans pass after correcting a helper’s executable mode and explicitly labeling public SHA-256 pins. The build-only temporary filesystem resolved the full image’s journal-creation failure. The repeated build and hosted CI then reached the unchanged created-directory owner check, which requires UID/GID 1000 while the build was running as root. The complete test-profile correction is integrated and independently reviewed: installation stays root-owned, then the unchanged verifier runs as UID/GID 1000 with a private HOME and the original sibling-import entry. The corrected full build passed all 121 authenticated controls and all 247 API controls, then exposed Python 3.10 retaining `/./` in the standalone test module origin. Both standalone test entries now insert the absolute working directory with matching source pins; all 176 unchanged host controls pass. The corrected complete image build remains pending; production guards remain unchanged. The four maintained-source lint errors are corrected with exact source-pin updates; the complete pinned selector passes 701 Python, four JavaScript and two CSS files, and full local CI passes. No build failure is waived. PR #291's 12 permission alerts received independent review: two installers expose only hash-pinned public source and ten tests deliberately induce permission drift. Individual security-thread replies record the evidence; CodeQL passes without changing scanner rules or checks. Installed runtime acceptance remains pending. The separate public authentication installer passed independent review and 14 root controls after fixing duplicate-gate and finite-path admission defects; exact installed image parity is pending. PR #289's native/image gates and target crypto checks passed, and its normal CI-approved merge is synchronized. These checks do not prove authenticated operational launch.


## Reader and backup verification 2026-10-11

The following status snapshot was moved from the task list. The current task list supersedes its pending states.

## Current hold-release work (2026-10-10)

- T028: Complete the authenticated fixed recovery launch and integrate the independently reviewed child/host failure-cleanup fixes. The corrected child launcher passed independent source review and all 52 controls in the selected Python 3.10/OpenSSL runtime. The corrected frozen host successor passes all 118 root controls and independent authority and backup/terminal reviews. It retains the complete verified file state through descriptor retirement and refuses late changes to the comic source, target and both databases. The existing Docker-volume layout and earlier initialization faults are covered. The live reader check found active WAL/SHM files and exposed missing source/target projections in the real preflight owner join. The bounded correction passed independent review and 121 source/relocated checks, including genuine committed-WAL backup/restore, while preserving the native immutable-read journal guard. Its five-path image packaging update is integrated and independently reviewed. Combined image parity and installed packaging now pass; actual configured launch remains unproved. Passing fixture signatures or enabling flags does not authorize recovery.
- T026/T029: PR #291's reviewed installer and image are merged and pass hosted CI, installed module parity and the selected runtime controls. The Btrfs correction in PR #293 passes full local CI and all 191 standalone and 132 authenticated controls on actual Btrfs under Python 3.10, with no skips. Hosted CI exposed one stale README source pin; the corrected complete owning-source copy passes. Three permission alerts concern independently reviewed hostile disposable fixtures and are documented as used in tests. The corrected complete local image build and final installed-module check pass. All 12 hosted checks pass, and PR #293 is merged through the normal protected workflow. Actual configured recovery and deployment acceptance remain open. [Earlier gate evidence](evidence-history.md) retains the original failures and detailed receipts.
- T017/T027: Complete a fresh consistent 12-scope backup and isolated restore before rollout. The failed operation retained both 58,287-file copies (170.68 GB each), but returned no final acceptance. Metadata diagnosis isolates source drift to Mylar stack-pins; the original verifier exception was discarded. The successor freezes that checkout during a short capture and verifies source, fresh independent Btrfs clones and all 72 databases once. Its efficient helper passes 54 Python 3.10/Btrfs controls; the corrected driver passes independent review and 15 source, 15 relocated and 15 actual Python 3.10/Btrfs controls, including late-callback refusal. Concrete original-input/capacity/vector binding and actual full-state verification remain required. Mylar and Komga are stopped; the worker remains held.
- T026/T029 follow-up: Configuration-time cache cleanup could delete retained sources before publication admission. The correction defers that cleanup to verified owning operations. Its successor image and hosted CI caught stale Config source pins; all six scope copies, dependent callers and finite source maps are now updated together. Installed and upstream preflights preserve the hash checks, with the upstream check following the complete patch sequence. The corrected source passes 115 focused controls locally. Its complete successor image build passes, and the actual installed export matches all 101 runtime source pins under independent review. Both unchanged isolated editions pass: genuine discovery and all 16 import/recovery/rename/metadata cases, with no skips. Full local CI also passes after integrating current main. All 12 current-head hosted checks passed and PR #296 merged normally. Configured archive/worker/reader terminal acceptance remains required before rollout.
- T022/T029 reader follow-up: The actual 490 MiB Komga database exceeded a mistaken 256 MiB file limit. The streaming correction passes 68 canonical and 68 retained-fixture controls under the existing 1 GiB file bound. Actual read-only execution then crossed the separate 256 MiB encoded-row budget at 386,490 rows in 14.63 seconds; the row and deadline limits did not fail. Original database hashes, identities and stopped profiles remain unchanged. Compact typed row fingerprints now preserve all references within that budget; binding and editable native rows remain explicit. The actual 389,249-row database passes in 14.68 seconds, with original database hashes, identities and stopped profiles unchanged; 77 canonical and 77 retained-fixture controls pass independently. Installed-image and configured continuity acceptance remain open.
- T018/T019: Verify the current library, DDL recovery, restart and database integrity before releasing processing holds; retain one verified current backup and unresolved originals. No live recovery or hold release is accepted by the checks above.

PR #304 commit d40ad9c7a0f7c967781676322fd13c45e14c8dfd updates both reader copies and their complete source/cohort maps. Reader source SHA-256 is 05affe629333eb81c84180e012d7d497baca73c7c39ce8c877facfc2b2a165a9. Typed row hashes retain complete values and duplicate counts for reference tables. BOOK, MEDIA_PAGE, MEDIA, issues and annuals retain explicit rows for owning operations. Independent review found no caller or preservation-contract mismatch.

The original read-only 513,683,456-byte Komga database crossed the old encoded-row budget at 386,490 rows. The corrected observation checks all 389,249 rows, 53 tables and 135 schema entries in 14.68 seconds, or 17.07 seconds including detached-copy verification. The existing 1 GiB file limit, 256 MiB logical limit, one-million-row limit and 120-second deadline remain. Final checks confirm the original database hash and file identity, stopped-service profiles, absence of journals and empty scratch. This proves reader mechanics, not configured recovery authority.

Root independently ran 77 canonical and 77 retained-fixture controls. Source and relocated controls, complete source-pin checks, make validate and make ci-local pass. The initial default-interpreter validation lacked PyYAML. Repeating with the system interpreter passed without changing checks. Hosted CI is running on the current head. The complete candidate image build and installed runtime verification remain separate delivery checks.

The corrected backup driver passes 15 source, 15 relocated and 15 isolated Python 3.10/Btrfs controls. The actual full-state backup remains unaccepted. Independent review rejected two launcher drafts that could replace final expected metadata through mutable parser objects or validation callbacks. Those drafts were not executed against application state. A correction is under review. Private stopped-service baselines and exclusive backup parents are prepared. The private-directory preparation first exposed a Btrfs directory cursor issue, reproduced with disposable files and corrected by rewinding before its second listing. Failed private preparation directories are retained. No application data changed.

## Installed candidate and backup handoff 2026-10-11

The complete compact-reader candidate image builds successfully. Independent final-image export verifies all 177 expected source hashes, including the 171 canonical leaves. Installed cohort verification, genuine discovery and the unchanged 16 import/recovery/rename/metadata scenarios all pass without skips. The private root runtime receipt has SHA-256 `eaf293a5f1be02a3b0b3540f123ac50821b76af6d820963376d723d97b0874b3`. These 17 isolated cases complete the candidate runtime milestone; configured parent/provider/archive acceptance, hosted delivery and live recovery remain required. PR #304 still awaits its custom Mylar image check; existing required checks and security analysis pass.

The reviewed backup launcher preserves detached original inputs and refuses both previously found mutable-baseline faults. Actual execution first refused a filesystem-ID list/tuple mismatch before copying. The narrow representation correction passes nine source and nine relocated controls and independent review. A fresh invocation then stops before its first stage: read-only reproduction identifies the owning helper's `not-quiesced` rejection. The driver passed the original deliberately disabled planning scope to helper validation before its current stopped-profile check. Correcting this ordering must derive quiescence from genuine current observations and retain the helper guard, original plan and final source closure. Neither invocation grants backup acceptance. Their private partial evidence remains retained; Mylar and Komga remain stopped and the worker held.

## Reviewed launcher and source delivery 2026-10-11

PR #304 merged normally at `2af9cdd2805583a8ba8491d62f5cd5d1662ef50a` after all 12 exact-head hosted checks passed, with no unresolved review threads or observed scoped security alerts. The documentation branch incorporates that main revision. The primary checkout stays unchanged during backup verification: its Komga normalizer source remains part of the long-lived original-source check even after the short Mylar snapshot.

The launcher successor derives a separate executable scope only after genuine stopped-profile checks and exclusive acquisition of the original writer lock. The disabled plan remains unchanged. Independent review found one further same-valued quiescence-object replacement accepted during a callback. The final correction captures its original identity first and refuses that exact fault. Driver 20+20 and wrapper 10+10 source/relocated controls pass without skips; independent review repeats 20+10 and all four original fault refusals. The selected Python 3.10 image passes the predecessor's unchanged 19+10 cases, then the final two focused disabled-plan and identity-fault cases; disposable Btrfs scratch is removed. These tests have no live application mounts.

Root verifies the complete source and review receipts, retains the previous failed directories, and binds exclusively new private paths. The actual full-vector estimate is 111,221,288 bytes within the unchanged 128 MiB bound. The fresh invocation returned held after its stopped-profile stage with `clone-original-birth`, before the current-stack snapshot completed. Source inspection identifies an unconditional same-device birth check in the ordinary-copy path; the actual source and backup disk are distinct filesystems. The correction must permit ordinary cross-filesystem copying while retaining same-device admission for FICLONE and all exclusive identity/preservation checks. Complete source/copy/database verification and the final receipt remain pending; no recovery, cleanup or processing release is accepted.

The cross-filesystem successor changes only the ordinary-copy device condition; FICLONE retains its same-device requirement. All original preservation/identity checks remain. Driver 23+23 and wrapper 10+10 source/relocated controls pass, and independent review repeats 23+10 plus all four original fault refusals. Three real two-filesystem cases pass in the selected Python 3.10 image: complete ordinary copying, refused cross-device reflinking and genuine 12-scope/72-database fixture adoption. No filesystem statistics are mocked. Root verifies the source and review receipts, retains the partial failed snapshot, binds new paths under the original private parents and launches one new invocation. Full-state backup acceptance remains pending.

## Published source and rollout preparation 2026-10-11

PR #305 merged the compact task lists and evidence histories after all required checks passed. GitHub and the authorized Gitea mirror both point to `d2889fc4246178d6536676993b2631ab2a7f2c4c`; merged-main validation also passes. Pages correctly did not redeploy for these specification-only changes. The primary checkout remains deliberately pinned while its Komga source is part of the current backup verification.

The actual trusted-main custom-image run for source merge `2af9cdd2805583a8ba8491d62f5cd5d1662ef50a` successfully published both Mylar tags to GHCR, with manifest index `sha256:b5ba2bf398f01bf12981ec7a0b309550608989402846953afcc0b0484c0afe3e`. Registry publication is distinct from the previously verified local candidate and does not grant configured recovery acceptance. The accepted worker target remains unchanged: all 35 production modules match its retained 679-control build, installed parity and passive alias evidence. No worker rebuild is required by the compact reader change.

The fresh backup and restore copies each contain 58,290 files, and the provisional manifest is written; final source/hash/database verification is still pending. Mylar and Komga remain stopped and the worker remains Created with no process. Source review rejects the historical selector/rollout helpers for current use: they rely on old backup/configuration bindings, and their original source closure needs correction. The corrected image-only selector passes independent review, seven source and seven relocated controls, and all seven controls in the selected Mylar Python runtime, with no live mounts or lifecycle commands. It retains original Compose files, parent/link identities and detached successor facts through final closure. Genuine held initialization, current configuration and source identities must still be proved after backup acceptance. No deployment, library recovery or processing-hold release is accepted by this checkpoint.

## Original-lock startup review 2026-10-11

The actual candidate entrypoint matches its configured source hash. Independent review follows configuration initialization, the held return, directory checks, optional backup, web initialization, version loading and scheduler startup. Holding the original exclusive Writer descriptor through the authenticated primary-key health response can keep this daemon initialized without native catalog/background-media startup. After descriptor release, that same daemon still requires a restart before ordinary processing. This conditional source result does not establish a durable hold across an unobserved restart.

Configuration initialization still rewrites configuration and can create directories; optional automatic backup, missing TLS files and release metadata have separate effects. The rollout must check those original settings and paths, installed image metadata and completed callbacks, then verify the actual same-incarnation held response and final authority census. No settings override or invented pending marker supplies this evidence. The private source-review manifest has SHA-256 `761f00f30bce126ccfddce3de65151fe41e2fec44c24f42d62e3260793af031c`.

The fresh backup contains 170,675,045,253 bytes across 12 scopes and 72 databases. Its verifier has advanced to the isolated restore copy; final full-state verification is not accepted. An independent task audit confirms 55 of 71 tasks complete: T026/T029 still need connected owning-factory/archive consumption evidence, and feature 024 T015 retains its explicit installation dependencies. Existing successful source/image/delivery checks are preserved; they do not close those remaining criteria.

The corrected startup diagnostic passes all 11 controls in the selected Python 3.10 image, preserving the original ten controls and adding the actual nested-DATA blank-cache regression. The actual stopped-service preflight completes with all selected configuration checks passing: current schema, disabled automatic backup/provider clear/update/browser launch, existing startup directories, API, settings, disabled TLS or configured-file existence, and original Writer identity. It makes no configuration or lifecycle changes. The private actual receipt has SHA-256 `d865d729035a0aa5a32cea3883d3f981c03e771cd3859d7f5e071517bc1e8735`. Installed argv/version completion and original-lock same-process startup remain separate rollout checks; this result does not release processing.

The disabled held-startup launcher now derives the expected successor environment, image labels and Compose identity from original selected Compose inputs and accepted image metadata. It uses the installed Python interpreter and requires the same daemon’s completed Main startup signal registration before Writer unlock. Review caught a rejected observation surviving the timeout; the corrected exact-loop regression refuses both an incomplete startup and foreign argv. All 12 controls pass in the selected native image without skips, and independent source review is clear. The frozen proposal manifest has SHA-256 `df0ab8e99e34ea567e1366e02037bf4f777b45f57a68f7508fb872d2163d18f7`. Operational bindings remain unset; these controls do not establish configured recreation, same-process hold/census or backup acceptance.

The fresh invocation subsequently exits held during verification with `verifier-OSError`. The provisional manifest remains unchanged; verification, original-vector and adoption-proof receipts are absent. Storage checks show ample backup and scratch capacity. The kernel records genuine checksum failures on the backup filesystem, with scratch on a distinct filesystem. Exact metadata mapping and bounded readability diagnosis are underway. The failure is retained and does not authorize rollout, backup cleanup or processing release; full-library verification is not being repeated merely to recover the discarded exception details.

Bounded diagnosis identifies exactly one unreadable 7,173-byte worker-state file in the isolated restore copy. The original source, fresh backup counterpart and retained seed counterpart are readable and match the unchanged manifest checksum; the original source identity also matches its recorded stamp. The corrupt restore inode remains untouched. A disabled one-file replacement proposal preserves that inode in quarantine and streams a distinct replacement from the original source; independent review and selected-runtime checks are underway. Replacement alone cannot accept the backup: the unchanged full source, copy and 72-database verification must succeed before rollout.

The corrected replacement passes independent source review and all 13 controls in the selected Python 3.10 image, including a genuine 12-scope/72-SQLite fixture. It retains the actual original Writer lock around current stopped-profile and writable-mount checks. Review fixes complete directory enumeration and late cleanup drift; the unchanged failure probes now refuse. The scoped actual invocation exits successfully: only the restore-copy file is replaced by ordinary streaming, and the original corrupt inode is retained in a private quarantine. Its original manifest and live source are preserved. The repair receipt explicitly leaves backup and application recovery unaccepted. A single unchanged full source/copy/database verification is the next gate.

The first post-repair launch stops at input validation before scanning archives: it bound the repair helper copy where the original backup request requires its separate frozen helper copy. Both original sources have identical unchanged bytes. The correction preserves both original identities and the immutable repair receipt; full verification remains pending.

The corrected launcher passes independent review and all ten selected-runtime controls, including distinct original helper copies, changed bytes, stale identity and late cleanup refusal. It retains both originals while invoking the unchanged verifier once. A fresh, separately bound invocation is now running with actual I/O observed; complete source/copy/database verification and final closure are still pending. No images are selected and processing holds remain in place.

The rollout-input preparation refuses the retained startup preflight because its original baseline checksum differs from the current backup request. A fresh invocation of the unchanged reviewed diagnostic uses that exact original plan and baseline, completes its original-file closure, and passes the strict baseline/configuration/Writer joins. The resulting private references include both original Compose/environment sets, accepted image metadata and 178 retained source hashes. Backup, selected-environment, API, argv and output execution fields remain unset; no selector or service action occurs. The earlier refused preparation and preflight remain preserved.

API/argv preparation refuses the host-loopback assumption because the original baseline has no published host port. The disabled correction observes authenticated API responses through Docker exec in the exact candidate, using its internal loopback configuration without changing network or privilege settings. The original 12 controls remain unchanged; all 18 source and relocated controls pass, and independent source review is clear. Selected-runtime execution and refreshed input preparation await operator authentication; neither has returned an acceptance result. Full source/copy/database verification remains running without final acceptance, and configured startup and processing release remain pending.

The first selected-runtime attempt fails before discovering any test because its private source directory is unreadable under the isolated runtime profile. The unchanged retry uses a verified public-source projection and passes all 18 controls in the accepted native Python 3.10 image without skips. Input preparation and independent original-reference checks also pass, deriving the internal API observation and ordinary argv from the original configuration. Existing wrapper, selector and other held references remain unchanged; backup, selected-environment and output execution fields remain unset. These results accept the controls and prepared inputs only, not a daemon, configured startup or rollout. Full source/copy/database verification remains active without final acceptance.

### Outer-backup prerequisite accepted — 2026-10-12

The corrected full verifier exits successfully after complete original-source and isolated-copy content checks and all 72 database checks. Independent verification joins the current original receipt, vectors and raw metadata across all 12 scopes, covering 58,290 files, 13,857 directories and two symlinks. The scoped repair and quarantined corrupt inode remain preserved; the original failed invocation is permanently unaccepted. This accepts the outer-backup prerequisite only, with no selector, service, recovery or processing-release action.

Selector preparation catches the original Docker executable exceeding its 16 MiB source-read bound before invoking the selector. The narrow successor raises only the original-file read bound to 32 MiB, retaining the Compose limit and original seven controls; all nine controls pass in the selected runtime without skips and independent review is clear. The initial discovery failure and corrected runner are retained without source or assertion changes. Finite personalization has independent source approval, but its selected-runtime check and actual input binding remain pending. Held startup, reader and NFS recovery, same-process custody and current-library acceptance still remain.

The first actual selector passes prepared-input checks and the three stopped-service profiles, then refuses strict Compose comparison because the worker inherits an unexpected image-selection environment entry. Owned rollback restores both original environment files, independently verified by their checksums. The narrow projection correction passes independent review and all 11 selected-runtime controls, preserving the original nine and adding known-projection and unrelated-value rollback checks. After a fresh binding refusal before writes, the finite correction accepts only three operation-owned rollback records. The actual retry sets both reviewed image settings and passes full strict comparison, without container changes or lifecycle actions.

Held startup first refuses equivalent null and empty-string user representations before recreation. The narrow normalization correction passes 20 selected-runtime controls. Its actual retry recreates the native candidate, then refuses a candidate-profile comparison and stops it through the guarded failure path. Read-only checks isolate the Compose image label: platform inspection matches the accepted configuration and root filesystem and supplies the runnable-manifest label. The image-index pin and platform-manifest identity remain distinct; the actual container retains the accepted image pin.

The platform-label correction passes all 22 launcher controls in the selected native Python 3.10 runtime without skips. Fresh capture verifies the stopped native successor against the complete expected security profile derived from pinned platform metadata, with both other service profiles unchanged. Independent configuration checks match the original backup bytes exactly; only modification and change timestamps differ, while the original Writer identity remains unchanged. That metadata difference is retained rather than silently refreshed. Fresh preflight and proof-linked configuration/baseline bindings remain pending; held startup, reader/provider/archive recovery, NFS and current-library acceptance are still incomplete.

The next actual held-startup attempt passes fresh preflight and proof-linked preparation, then recreates the native candidate and refuses HostConfig bind-list order. Guarded cleanup stops it; read-only checks confirm the same two unique bind strings and all other static fields. The narrow correction sorts only those unique strings while retaining shape and duplicate refusal. The first runtime invocation runs only 12 controls; the corrected combined invocation passes all 24 without skips. Fresh baseline and manifest-backed configuration comparison preserve exact bytes and setting values, with the original Writer unchanged and timestamp drift explicitly retained.

The V7 actual parent passes recreation, full profile comparison and primary API health, but refuses its Docker version guard: current/latest versions match the pinned release while commits-behind is null, as the original Docker-loading source specifies. The bounded invocation times out and confirms its candidate stopped, without a startup receipt. The verifier-only correction passes independent review and all 25 selected-native runtime controls without skips; no product source or image changes.

The next fresh stopped-baseline capture and configuration comparison pass again, with exact configuration bytes, no changed, removed or added settings, and the original Writer unchanged. The finite V8 parent bindings pass independent review. Its actual invocation completes successfully through the same daemon PID, Main signal-registration witness, 178 installed source checks, complete expected profile and authenticated health/version observations. The parent confirms writer-busy before unlocking the same original Writer FD, then startup-restart-required with a nonempty census. Fresh read-only confirmation repeats the same held health and nonempty census against the original startup receipt and verifies both old reader/worker profiles unchanged. Native startup is accepted in its held state; Mylar is running, Komga remains stopped and the worker remains created. Reader/worker profile preparation, genuine reader/provider/archive recovery, NFS and current-library acceptance remain pending. Processing is not released.

The user requested normal Mylar availability alongside the remaining recovery work. Source review identifies the remaining hold as the startup-only latch; the owning README requires releasing the external deployment guard before normal startup, while pending media recovery continues to require explicit authenticated review. Normal availability is separated from the unfinished bulk, experimental-worker and archive-recovery gates, with no flag overrides or automatic recovery replay.

The actual normal restart completes with API READY, the Main signal-registration witness and two identical observations of all 178 installed sources bound to the new container/daemon identity and start time. The container profile is unchanged, existing ComicID/IssueID associations are preserved with new additions allowed, and SQLite integrity passes. The internal home request returns HTTP 200 HTML/login content without a processing-hold message or server error; authenticated DDL pages are not accepted by that check. Mylar is available, while the old reader remains stopped and the experimental worker remains created without a running process. Bulk, archive recovery and complete current-library acceptance remain pending.

Normal Mylar remains READY. The current DDL snapshot contains 97 DDL rows, including seven completed packs: 26 marked Completed, six Failed and 65 NZB handoffs, all queued. The source probe records 26 captures; 25 sources are available with matching recorded stamps and distinct path digests. Content identity is not used to deduplicate retained generations. These status, queue and capture observations do not accept final publication, reader continuity or retained recovery. Fresh worker verification accepts its image and complete expected profile in Created state with no running process; the worker stays inactive and Komga remains stopped. Experimental/bulk admission and complete current-library acceptance remain pending.

The compatible reader starts successfully under its owning guards and is healthy in the same original container. All schema/table fingerprints match the verified backup before and after startup; normal Mylar remains READY and the prepared worker remains created and inactive. Earlier strict start-profile refusals are retained. The narrow network correction equates absent and null entries only for unpublished exposed ports, while refusing host bindings and preserving other profile, backup and data guards.

The authenticated DDL page is available. Its current Remove confirmation is opened and cancelled without performing a deletion. Neither this UI observation nor compatible reader startup accepts the pending archive repair, retained handoff, NFS canary or complete current-library work. Experimental/bulk admission remains held.

The DDL recovery batch adds an exact capture/DDL maintenance pass without bulk conversion, naming or library scans. Downloader status and external-provider failure handling now retain the original row's provenance and skip removed or replaced rows; existing ownerless Failed records are not assigned guessed owners. Integrated source controls pass 696 checks with zero skips, and the custom image gate now checks the installed status guard before any temporary installer repair. A separate 20-check regression verifies metadata policy reads under actual native admission and continued idle-consumer liveness after refusal. Repository validation and full local CI also pass. Matching images and live recovery remain pending. The fresh queue observation reports enabled post-processing without a live worker and 12,018 waiting search entries; another chat confirms it has not added or held that search work.


Hosted checks expose unused imports and a selected-receipt fixture that assumes directory enumeration order. The narrow corrections preserve evaluation and exception behavior and bind the fixture response to the selected token, retaining its ownership, fence and no-replay assertions. Production receipt handling is unchanged. Local privileged verification and deployment are paused at the operator’s request; hosted checks remain mandatory.


## Ready Writer-session source proposal 2026-10-12

The private coherent proposal is based on the merged PR #318 source tree at `ac4364a0870400e56fce9143828706f5125c4b44`; it is separate from that delivery and remains unmerged. CI's delivery checkpoint owns the 14 hosted results, slow native gate, worker gate and pending trusted-main publication. This source milestone does not establish publication or deployment of the Ready successor.

The existing parent pipe now seals the complete prelease census and exact local Controller/Writer paths. Enter precedes acquisition; held challenges and idle checks make no Docker or API calls. ACK remains inside the original Writer, with a pending terminal reply granting no rights. One post-ACK leave follows release, before independent terminal/worker checks and natural exit. Failed acquisition, original drift and uncertain responses refuse without replay or refreshed expectations. Normal Mylar availability is separate from these unfinished recovery gates.

The coherent source gate passes 758 cases across 26 top-level suites, including the 15 connected Writer-session controls, plus the separately declared 591 recovery checks. The latter count is the existing 588 plus three complete-census NFS cases (35 to 38); connected15 is not part of that increase. These are executed case counts, not unique coverage totals. Full authenticated132 and copied-source6 controls also pass without skips. Relocated exact-source checks pass connected15, installer23, copy6, NFS38 and the eight owning runner controls, which execute recovery591. Ruff checks all 39 changed Python paths; Python3.10 syntax checks pass. Actual selected-image/runtime execution is not claimed.

The historical V3/V4 record incorrectly said six configured-scope aliases matched; five were updated while the differently named fixture_scope retained V2. The coherent proposal updates that sixth alias and verifies all six equal. The earlier NFS receipt label said38 but its log proved eight archive-readiness cases; the actual amended NFS38 suite is now separately run and retained. The first full reader-gate attempt fails only its historical588 count assertion after genuinely executing591; that failure is preserved and the factual count correction passes. All earlier source, callback, local and operational refusals remain evidence rather than waived results.

The coherent manifest is `0b8e04bea3a3425b9c2f6ecee84ca6522899b9973866bff41335ed9f03491050`. Relocated evidence is `45e64e81bf1878f2198fd75b460c6fac757acd17966c9f1664ca35151908a72d`. Independent source review passes, including the complete alias, map, installer and image-gate propagation. Its receipt is `0271c280c9ae47b518abffa0af000debd7b2ada009fbceaf28903e524ae569ff`. Root verifies all 50 original preimages and successors against the merged tree. Repository validation and full local CI pass on the integrated source.

The source maps cover 74 SDK modules, 25 cohort modules, 178 recovery files and the existing API, standalone and authentication controls. Worker production and Docker runtime contracts are unchanged. T022/T026/T029 and feature 024 T012/T015 still require matching installed origins, original configured inputs and actual archive, worker, NFS and reader terminal proof. Naming, bulk and current-library acceptance remain open. New pkexec calls, deployment and live mutations remain paused.


## Negative-five Writer-session source successor 2026-10-12

The bounded T022 audit finds a separate source gap after the archive Ready proposal: negative-five purpose_existing entered a plain Writer hold, then scope revalidation sent a normal parent challenge with remote observations under that lease. The actual owning caller, real configured-scope factory, two pipes, ordinary Writer and initialized SQLite reproduce the refusal. The source successor uses the existing announced session only for reviewed negative-five version9 prepare/execute. Backup, verify-terminal, foreign plan kind/version and stale SQLite census refuse. Archive10–12 phase behavior is unchanged. The negative provider's main report, watch and ordinary ACK remain after release; no archive pending-terminal exchange or terminal/resume right is added.

NativeBatch retains its original census and records. Existing forward/rollback terminal proofs require that same ledger, and the coordinator consumes only the genuine complete typed terminal. The successor neither refreshes census nor changes the underlying SQL, marker, rollback or custody guards. Nine connected controls, all 33 unchanged provider test methods, the 86 parent and 15 archive connected controls, and existing forward27/rollback19/coordinator29 checks pass without skips on host and affected relocated copies. Their SDK/reader boundary projections do not establish installed factories or configured aggregate execution. A direct public0644 fixture attempt fails the unchanged frozen-kernel guard; the existing owning runner's private0600 kernel profile passes with identical source bytes. That failure is retained.

Independent semantic and coherent source reviews accept the successor. Its complete source-map closure and mandatory nine-control gate pass; the full host gate executes 767 top-level and 591 recovery controls without skips, with relocated results retained separately. The source bundle is prepared and independently reviewed with IMAGE and ACTUAL inputs unset; it establishes no operational authority. The prior proposal manifest is `68f98c5d933915d69c099f9cf964e711f75b7a3f4d65f661c1e9dd1bcb29e42d`, relocated evidence is `1e8a449a0d35a7eccafd8ed368243ffbfc3be2d3c153cda0c8a3eea1f572061a`, and independent review is `139a33b5961873c5bcaede9577c07e74889445585b96083e247cf2dfed8a5ca5`. Current source parent/provider originals are e43d/9515; historical bindings cannot supply current authority. Selected-image Python3.10, current installed origins/cohort, original five-owner/reader-pair/eleven-row inputs, NFS, backup/restore and actual terminal acceptance remain unproved. No whole task checkbox or new ID changes. Normal Mylar and reader availability remain separate; new pkexec calls and live mutation stay paused.

### DDL source and image delivery — 2026-10-12

PR #318 merges after all 14 hosted checks pass. The full native image gate and all 689 worker controls pass without skips; trusted-main validation, security, Pages and both image publications also pass. GitHub and its authorized mirror agree. Earlier local attempts remain failed and preserved: immutable context modes blocked fault-fixture writes, and the selected-receipt fixture assumed directory order. Its correction retains ownership, fence and no-replay assertions. Local privileged verification and deployment remain paused; no live queue recovery or experimental acceptance is claimed.

PR #319 merges after all 13 hosted checks pass, including the complete native image build, with exact reviewed-to-merged tree equality. Trusted-main publication is still pending. This is source and hosted-image acceptance; matching installed origins and configured archive, worker, NFS and terminal execution remain open.

## Published successor image preparation — 2026-10-12

Privileged access resumed. Read-only inspection confirms Mylar and Komga are running and the normalizer remains Created with no process. The final native image from PR #320 (`ca13a184…`) and worker image from PR #318 (`49e9103d…`) downloaded by exact published digest. Native identity inspection confirms Linux/amd64 and the exact digest. Worker identity inspection and the current installed-source/runtime checks remain pending. No live service was changed. Earlier image proofs are historical and do not establish acceptance of these successors.

The worker identity check subsequently passed with the exact digest and Linux/amd64 platform. The combined isolated native check stopped in the Docker absent-name preflight before creating its first test container; no installed-source or runtime acceptance was produced. The authenticated diagnostic confirmed an absent container with `Error response from daemon: No such container`, a response the preflight did not recognize. A narrow parser correction is being prepared. The retained failed run is distinct from earlier successful candidate checks.

The corrected preflight permitted the actual published native image check to run. Installed three-file parity, SDK/cohort verification, the current 101 native plus three auth export joins, canonical source-map joins and the strict discovery import case passed. The retained 16-case batch stopped at its nested harness source-pin check (`reviewed T028 source pin mismatch`) before executing that retained case. This is partial isolated proof only: the remaining cases, live deployment and configured recovery stay open. The original failed run is retained; reconcile its expected pins with the verified export before rerunning the affected batch.

## Current candidate runtime acceptance — 2026-10-12

The published native candidate passed the 16-case rerun after the expected-pin-only fixture correction; the strict discovery case had already passed. A separate postcheck joined the complete failed and successful fixture inventories to the original published verification manifest, current 104 export and 101 runtime source joins, ordered zero exits, original output hashes and receipt. It did not repeat the 16 cases or relabel the V2 execution as a V3 run. Actual counts are 177 Python leaves and 171 canonical map entries.

The same batch then ran all 25 distinct rollback/combined-rollout controls in the actual candidate Python 3.10 image: 18 rollback and seven combined controls, zero skips. The private receipt is retained outside Git. The wrapper manifest is `c6a10666…`; unit output hashes are `f8fe47d2…` and `b0bfb24f…`. The isolated tests use real disposable backup/SQLite/restoration fixtures with projected Docker/API observations. No live application mounts or deployment were involved. Fresh baseline binding, consistent backup and isolated restore, live rollout and configured recovery remain required; whole task checkboxes remain open.

## Cache capacity and rollout preparation — 2026-10-12

The candidate image passes the paired preparation/API timeout compatibility check in its actual Python 3.10 runtime. Fresh read-only capture reaches the scope census, then refuses an internal symlink in the worker’s historical backup tree; no service stop or deployment occurred. The correction must prove target coverage without relaxing external-link refusal. Storage inspection identifies 45.7 GB in Mylar’s retained download cache and approximately 2.3 GB in operation-created reader-state directories. Repeated pack names and equal sizes are not content-equivalence or import evidence. Docker reports 30.4 GB reclaimable build cache; physical reclaim and artifact ownership remain unverified. No cleanup occurred. New FR-016/T030 records prevention and its acceptance checks; existing source/image/live milestones remain distinct.

The cache allocation core now passes nine independent host controls with real disposable files and process-level lock contention. It counts retained and temporary entries through no-follow directory descriptors, retains original lock/destination identities, and allows zero-allocation recovery even above the configured budget. Streaming/extraction integration is in progress; alternate provider write paths remain under review, so total-provider coverage is not yet claimed. The corrected internal-link coverage passes six source controls and independent review, including the retained alias/parent traversal failure. Actual candidate-runtime tests and the new read-only cache inventory are waiting for desktop authentication; no rollout or deletion has run.

The integrated cache proposal passes the original nine transfer controls, 26 transfer/extraction controls, 15 actual provider-function controls and four SQLite queue-generation controls in host verification. Mega now reserves before temporary-file creation and plaintext writes; its stage belongs inside the measured cache and promotion cannot overwrite an existing destination. Exact complete extractions are reused; altered or incomplete companions remain retained for review. Late Remove, handoff and generation changes refuse capacity requeue. Source-map and selected-image verification remain pending. The native-only backup proposal now covers native config/state and rollout pins, preserving all three service profiles and refusing actual mount overlap with worker state; 27 host controls and independent review pass. A fresh worker-state backup remains required before worker update/start/canary. No live rollout, deletion or whole-task completion is claimed.

Independent review of the third cache proposal verifies all 650 source pins and passes 36 new plus nine original controls without skips. It closes the reproduced final-transfer row resurrection after Remove and classifies incomplete extraction as retained review. Those results do not accept later installer/provider changes: final integration must use the actual nested downloader package and guard metadata updates from the provider entry generation, including lookup/progress callbacks. The retained failing probes remain evidence. Configurable defaults are 20 GiB cache capacity and 5 GiB free-space reserve; no live guard, cache deletion or complete T030 acceptance is claimed.

The final cache source proposal (`bb14737b…`) seals 649 pins and 33 source paths. Root verifies every hash/mode, integrates the installed-parity and focused image gate, and passes 58 focused test executions (including inherited controls). Original-entry metadata updates cover Requests/GetComics and the actual nested MediaFire, Pixeldrain and Mega providers; real SQLite cases exercise lookup/progress Remove and NZB handoff. Network/crypto and test free-space capacity are explicit projections. Independent final review and reconciliation with the separately merged UI work are pending; no selected image or live acceptance is implied.

Final independent source review (`35ce070c…`) verifies all 649 pins and passes 65 distinct controls with zero skips: 40 new, nine original transfer, 11 original status, three HTTP and two original exhaustion. Original negative assertions remain unchanged; the positive HTTP test now checks the actual four-byte SQL metadata result. Root verifies all 11 review artifact hashes. Integration advances the owned cache branch to the separately merged UI main, preserving both sets of image tests; two generated hash lists require regeneration from the combined source. No live or installed acceptance is claimed.

Combined UI/cache reconciliation (`0d2a018f…`) preserves all cache semantic leaves and both image-test families. All finite source maps and SDK/cohort/recovery joins close after regeneration, including the merged UI import-problems pin. Affected copy, cohort, provider and UI checks pass; the optional target-runtime family remains unproved outside its backend. Full local/hosted CI and matching-image/live acceptance remain required.

Full local CI passes on the combined candidate with installed workflow/Dockerfile/YAML tools and no optional-tool skips. The earlier restricted-PATH attempt failed tool discovery and is retained as a failed attempt. Maintained changed Python passes focused Ruff; staged scanning and the 554-commit history scan find no secrets. The whitespace exception preserves only the six hash-verified original `.py.txt` fixtures, following the existing imported-fixture policy; production/test files keep normal checks. Hosted image/CI and configured live acceptance remain pending.

A separate preservation review (`85ad5a86…`) holds V5 despite its passing controls: destination ancestor collection overwrites previously captured expectations, and final transfer promotion can replace a newly appeared foreign file. Both reproduce with genuine disposable files/SQLite; the exact failing probes are retained. The narrow successor must preserve original ancestor expectations and promote without overwrite while retaining foreign and partial bytes. The combined candidate is committed locally but not pushed or deployed; source delivery remains blocked on these corrections.

The narrow successor (`c460b253…`) changes six source/test/map paths and closes both retained preservation failures. Independent re-review (`04c709ce…`) verifies 651 pins, preserves all 32 original focused test method ASTs and passes 43 discovery executions without skips (including inherited controls). Untouched original probe portions now refuse both faults; the new collision control preserves the foreign final and complete verified partial and leaves native metadata incomplete. Root verifies source and review hashes. Full local CI passes on the corrected tree, including the 555-commit history scan. Matching installed/live acceptance remains separate.

PR #323’s first hosted run refuses the archived queue fixture’s missing cache helpers and the installed Config pin changed by the new cache preferences. The correction copies the two exact owning helpers, derives the new pin from the checked original Config and actual ordered transformations, and closes the dependent scope/source maps without changing allocation, metadata-CAS or preservation behavior. Nine configuration, 35 configured-scope, four parent-admission and two cohort-import controls pass locally. Fresh full CI and installed-image gates remain required; no live update or whole-task completion is claimed.

The same hosted run reports CodeQL alerts 150/151 on newly created HTTP partials and extracted cache members. Their permissions are tightened to owner-only at creation, matching the existing private staging contract; resumed originals and final library permission policy are unchanged. A genuine file-mode regression preserves the original input permissions while checking the new transfer and extraction payloads. Thirty-three affected transfer/provider controls pass without skips; no scanner suppression or live permission change is used. Both narrow corrections pass independent review, the authenticated host suite passes 132 controls without skips, and full local CI passes on the stable corrected source. The earlier local run that observed source updates during its checks remains retained; fresh hosted verification is still required.

The corrected PR #323 head passes hosted source lint and repository validation, and the installed Config preflight succeeds. Its image gate then identifies a build-context omission: the new cache fixtures were excluded by the existing default-deny Docker rules. The correction admits only their exact public directory, six provenance-checked source leaves and provenance JSON; runtime configuration remains excluded and all test assertions are unchanged. Matching-image and live acceptance remain pending.

The upstream mechanical image gate then identifies a missing disposable sibling library: temporary installer replay copied the native modules but not their actual Mega dependency. The verifier now copies that exact image-owned library into the private sibling layout before applying patches. A regression executes the owning copy statements with only the fixed public source path projected, then proves real installation; all five installer controls pass. Installed-source verification remains separate and its guards are unchanged. Fresh hosted image verification is still pending.

The next hosted mechanical gate reaches the retained native HTTP resume case and identifies its obsolete mock database/module setup. The fixture now uses real SQLite and the owning original-row generation helpers, preserving every Range, byte and partial-file assertion and adding the real Queued-state check. Only disposable free capacity is projected. Three resume controls and the existing late Remove/NZB preservation controls pass; production transfer guards are unchanged and fresh hosted image completion remains pending.

The following mechanical gate exposes a caller mismatch: the new self-contained cache suites received a native-source path that unittest treated as a test name. The verifier now omits that argument for those four suites while retaining it for every existing source-consuming suite. An owning-loop regression checks the exact command routing; test bodies and production guards are unchanged. Fresh hosted image completion remains pending.

After the cache suites pass, the mechanical gate identifies the missing real cache validator in two retained Config caller fixtures. Both fixture namespaces now expose the owning validator; existing setting validation, pre-mutation, rendering and persistence assertions are unchanged. Focused Config checks pass, while full template coverage remains with the hosted image gate because the retained source export omits templates. No production source or acceptance guard changes.

### Cache guard delivery — 2026-10-13

PR #323 delivers cache growth prevention through a normal squash merge after all 12 required hosted checks pass, including the full native and upstream Mylar gates. Independent reviews and focused regressions preserve the original queue generation, transfer bytes, permissions of resumed originals and setting-validation assertions across the corrected source pins, disposable library layout and fixture callers. Main validation, CodeQL, source lint, Pages and trusted-main image publication all pass; the source-specific GHCR tag is independently confirmed. GitHub and the authorized Gitea mirror match. Downloaded-image metadata checks are complete; matching runtime and configured live cache verification remain pending, so T030 stays open and the whole-task markers are unchanged.

### Cache guard runtime verification — 2026-10-13

The exact published native image `9ce03212…` passes recursive source parity for 206 Python leaves, installed-cohort verification and the original export joins. The initial helper batch stops before tests because the disposable layout lacks the expected preparation/binding paths. A dependency audit also identifies the omitted Scout source before rerun. The reviewed correction adds only the exact public helper copies/mount and moves original ancestor capture before the first reads; test bodies and owned-container cleanup stay unchanged. The failed batch and its source-capture ordering limitation remain recorded.

The corrected downstream batch completes all 33 helper checks, 25 rollback checks and 17 import/recovery cases without skips. It retains the three successful source stages rather than repeating them. Root execution exits successfully with private proof `gnuqxnxf`; independent source review `2ea41dda…` and the frozen batch `d544aba7…` are verified. These isolated results permit preparation of the native-only update, but do not accept live retained-generation recovery, worker processing or reader/archive terminal handoff. T022/T026–T030 and the linked 022/024 acceptance tasks retain their remaining conditions.

The separate read-only build-cache inventory initially reports no candidates because Docker renders last-use dates as elapsed text. The reviewed parser correction handles those dates and refuses fallback to an older creation date when a nonempty last-use value is unknown. The fresh inventory identifies 224 eligible records with an estimated 5.7 GB total, preserving active builds, shared layers, required image references and their ancestors. No cache had been deleted at that inventory checkpoint. The subsequent exact-ID prune command returns success but reports zero bytes reclaimed; the eligible-record inventory is unchanged. The subsequent corrected exact-ID cleanup succeeds: Docker reports 4.582 GB reclaimed and eligible records fall from 224 to 108. No service lifecycle changes occur. Filesystem usage remains at 98%, so the Docker accounting result does not establish a corresponding physical space gain.

The exact published worker successor also passes all 689 tests without skips through the unchanged Dockerfile discovery command after all 36 installed module hashes, import origins and existing passive type/disabled-rights checks pass. This accepts isolated worker verification without claiming configured factories, native/worker terminal consumption or live worker activation.

Fresh read-only native-update capture also passes. This prepares the scoped native-only backup inputs; it does not accept a new backup, update or configured recovery operation.

The first native-update attempt refuses before creating output or changing service lifecycle. A read-only execution of the exact prefix passes, leaving the cause unidentified. After authentication, the retry with observability around the original main entry runs the scoped backup. A read-only observation confirms the original native image is stopped with no process, Komga remains healthy on the same process and the worker is inactive. The operation output exists but is still empty; no backup, native update or live recovery is accepted.

## Native backup refusal and original-image restart 2026-10-13

The scoped native-only backup ends with `source-changed`; no new backup or candidate deployment is accepted. Independent recovery inspection verifies original-image API READY, version response, completed Main startup, two identical installed-source observations, preserved captured catalog identifiers and SQLite integrity. The exact later pre-operation catalog was memory-only, so that narrower comparison remains unavailable. Docker health is unhealthy: recent checks report post-processing and DDL workers down plus stalled processing. The UI chat reports post-processing was already down before restart, which does not establish a restart regression. The reader remains healthy and unchanged, and the worker remains inactive; the earlier verified backup and failed-operation evidence remain retained. A read-only scope check finds rollout-pin files inside the primary checkout and native configuration outside it. Observed checkout changes could alter metadata, but the first failed scope and cause are not identified; the next affected backup window will keep that checkout stationary.

A prospective reflink helper passes all 67 controls without skips in actual Python 3.10 on Btrfs. It is not the helper used by the failed operation and has not been activated. Separately, PR #327 merges through normal hosted CI and publishes the UI successor. Source-compatibility review confirms that preservation, import, cache, pack, reader, naming and repair algorithms remain unchanged apart from presentation and required source-pin updates. Current-cohort binding and installed-source proof remain necessary before deployment; publication does not establish configured live acceptance.

The matching PR #327 worker image passes fresh identity/platform verification, all 36 installed-source and import-origin checks, passive factory and disabled-rights checks, and the complete 689-test suite without skips. It remains inactive. Live logs also confirm the prior image's historical post-processing idle-poll admission failure, already corrected in the published successor. A separate source reproduction proves that a transient Writer-busy startup probe can silently skip queue creation. The reviewed correction waits through the existing bounded Writer operation and revalidates admission before scheduling; all 62 startup controls and seven independent concurrent-Writer controls pass without skips. Hosted pinned-source verification exposes a no-op mutation in the new drift fixture when given already-corrected source. The fixture now constructs and checks the exact predecessor guard before mutation, retaining both old-guard and new-guard refusal assertions. Other publication holds remain unchanged. The correction is not deployed, and current queue liveness still needs verification after the repair rollout.

## Startup repair delivery and remaining installed integration 2026-10-13

PR #329 merges normally after all 12 required hosted checks pass. Its full custom-image gate executes all 62 startup tests without skips; the upstream-only gate's six optional backend skips remain separately qualified. The earlier registry-rate-limit failure is retained and its normal rerun passes. Merged-main validation, source lint and CodeQL pass, and GitHub, the authorized Gitea mirror and the clean primary checkout agree on the accepted merge. Trusted-main publication and actual downloaded-image verification are still pending at this checkpoint; no new live rollout is claimed.

The corrected backup/rollout helpers pass 46 isolated controls in Python 3.10. Their finite private selector snapshot avoids using the mutable checkout as the affected backup scope, while original selector bytes, attributes and link identity remain checked. Candidate acceptance requires every enabled queue alive and healthy Docker status. The corrected preparation joins the new proof's own source/export/runtime results and current recovery epoch rather than the prior image's report or startup epoch. Its nine predicate controls also pass in the previously verified Python 3.10 image; this is helper verification, not new-image or configured acceptance.

A source-led task audit distinguishes isolated installation from live acceptance. T022/T026 and feature 024 T012/T015 can close after genuine installed integration and current-image delivery, without waiting for live T017/T027. Existing passive factory checks and projected or synthetic terminal/ACK fixtures do not prove that integration. Actual installed lifecycle-parent/provider and archive-terminal production with factual native/worker consumption remain missing; the negative-five default deliberately keeps its parent pin unset until explicit reviewed configured installation. Two disjoint implementation lanes now address those gaps. T028 still requires a genuine configured retained generation and fresh owning catalog-to-ACK event. No whole-task marker changes at this checkpoint.

The configured cache-guard plan separates no-payload accounting/refusal from actual queue status and provider-attempt preservation, unresolved-capture deduplication and complete existing-extraction reuse. It does not authorize manufactured ACKs, destructive capacity canaries or deletion of retained originals. These configured observations remain pending.

The genuine installed-route trace identifies a concrete ownership seam before birth admission: the privileged parent emits mode-0600 controls into a root-private operation, while the selected child runs as UID 1000 and the installed reader checks require controls owned by that child. Existing projected fixtures did not exercise this root-to-child boundary. A finite private control projection with continuous original custody is being designed; changing permissions on root originals or replacing installed authority with test globals is not accepted. The ordinary startup repair remains independent.

Trusted-main publication completes for PR #329, and the actual host pull verifies index `3166b2ca…` and its linux/amd64 manifest `20d464d9…`. Docker's default image identity is the index; the platform-specific identity is the manifest. The OCI config digest is separate. Both isolated attempts pass current source parity, installed-cohort and public-export checks. The first stops before helper tests because two expected fixture aliases are absent. The second stops at container creation because Docker cannot create those file mount targets within the read-only fixture directory. Both failed attempts remain retained. The corrected prepopulated fixture directory preserves all original 1,029 resource hashes through a mechanical path remap and adds the two exact alias files. Independent review and root verification cover all 821 artifact files and 1,031 runtime resource pins. Original test bodies and acceptance assertions remain unchanged. The fresh actual-image batch then passes all nine stages: source/cohort/export checks, 33 legacy helper checks, 25 rollback checks, 46 current rollout controls, six producer controls and 17 import/recovery cases, with no skips. Private proof `gbuf0v8j` retains the original raw outputs and command records. This accepts isolated image verification; it does not accept configured recovery or deployment.

A fresh independent read-only recovery check again verifies original-image READY, source, catalog identifier preservation and SQLite integrity, with unchanged healthy Komga and inactive worker. Docker remains unhealthy. The existing comparison's explicit limitation remains: the exact later pre-operation catalog was not retained. Fresh disk checks still report 23 GiB available on `/mnt/apps` at 98% usage and 317 GiB available on the backup disk. No live backup or update is running at this checkpoint; whole-task markers remain unchanged.

## Active native rollout and owning integration 2026-10-13

Fresh capture verifies the current service profiles, configuration, API/catalog and finite selector snapshot. The reviewed reflink backup/rollout helper is now running against the published PR #329 image. Mylar is stopped for its consistent native backup and isolated restore; Komga remains healthy on the unchanged process and the experimental worker remains inactive. Captured affected state totals 48.73 GB across 22,140 entries. The operation manifest now exists, but its verification receipt is absent. The latest read-only process check observes an open restore file; no completed verification, rollout result or recovery result has been observed. Byte counters are progress evidence, not backup acceptance. The earlier failed operation and verified outer backup remain retained.

Independent review of the first owning projection proposal reproduces a returned-file identity defect: the supplied inode signature was discarded and recaptured. The stage2 successor rejects that exact unchanged wrong-inode probe before creating a root view, routes Parent/history and producer reciprocal reads through ownership-aware views, and adds an exclusive configured installer. Sixteen routing and five artifact controls pass without skips. Independent review holds stage2: routing inserted absent history/baseline/preparation proof members instead of requiring their original physical identities. Stage3 narrowly restores mandatory mapped physical proof membership before deriving a root view, refuses late proof-table mutation, and passes 19 routing and five artifact controls without skips. The earlier proposals remain retained on hold; the correction has source review but no connected installation or runtime acceptance. Actual configured installation, genuine Parent/Birth/pipe transitions and terminal consumption remain pending. The separate current-role NFS adapter preserves observed names, source-cohort identities and NFS4 requirements; its actual integration remains unverified. None of this experimental work blocks ordinary startup repair.

### Reviewed installer and host-lease implementation

Stage4 adds complete known-source SDK derivation, one prospective admission overlay and physical child-copy/empty-directory inputs for a host-owned lease. Independent review verifies all 175 known source rows against the owning expected map and the complete 100-leaf SDK set. All 22 routing and eight artifact controls pass without skips. These are source checks; root installation and effective canonical origins remain unexecuted. The tracked integration audit identifies explicit manifest/test registration, the new JSON build-context exception and leaf-first producer/loader pin updates before delivery. Host-only helpers do not justify changing base SDK or worker parity.

The host-lease successor keeps a real process, FIFO descriptors, Writer lock, container identity and epoch. Review catches a post-acknowledgement original-file change and an exited-epoch gap; root also identifies rejection of legitimate inherited image labels. The corrected successor passes 15 independent controls without skips, and the unchanged mutation probe now refuses. Existing labels remain part of the immutable full profile. Persistent Parent dialogue, child factory wiring, actual Docker/NFS and terminal/native/worker acceptance remain open.

### Ordinary repair rollout and independent observation

The guarded native operation exits successfully after backup and isolated restore verification. Its result records Mylar READY, every enabled processing queue alive, healthy Docker status, original catalog preservation and released deployment exclusion. Komga is unchanged and the experimental worker remains inactive. A separate read-only observer joins the manifest and verification receipts: 32 database checks, 21,873 files, 266 directories and one symlink. It confirms Mylar image `3166b2ca…` running healthy on PID 778848, unchanged healthy Komga and the inactive worker.

The final independent API/source/catalog observer stops with ValueError. Its sanitized output does not identify the failed assertion, so acceptance is pending exact diagnostic evidence. This observer performs no lifecycle changes. Preserve the successful driver result and the failed independent observation separately; do not treat either receipt/profile checks or an unexplained verifier refusal as final application-data acceptance. No whole-task marker changes follow from this partial rollout milestone.

The exact unchanged observer is rerun under a reviewed diagnostic wrapper that permits only fixed assertion names and public function/line output. It returns successfully: Mylar READY, all enabled queues alive, Docker healthy, all 206 installed source checks, original ordinary Main and catalog preservation against the verified stopped pre-update restore. Komga remains unchanged and the worker inactive. The private acceptance receipt is retained as `_xeboo9s`. The earlier opaque ValueError is retained and does not recur; its cause is unproven. No acceptance predicate or production source was changed for this rerun. Ordinary service acceptance and the coordination handoff are complete. Experimental Parent/NFS, archive terminal, selected retained DDL recovery, bulk and library acceptance remain open.

Stage5 receives independent narrow source review with 32 routing/installer controls, including generated loader execution. Its constructor runtime remains held for original source-copy custody and success-after-cleanup corrections. The child factory/real host-lease interface package passes all 31 independently rerun host controls without skips. This includes genuine UID-1000 FIFO exchange with the real helper process and Writer lock. Actual Parent launch, NFS, Birth and archive terminal/native/worker consumption remain unproved. The probe constructor and persistent Parent session are being integrated in separate coordinated lanes.

After the successful independent observation, authenticated UI checks report Post-processing returning the styled hold fallback and DDL table/preferences polling failing, while Activity and Import Problems load. Full normal-route availability is therefore unresolved. Source inspection finds that the HTTP startup guard probes publication authority with a nonblocking Writer lock and can report `writer-busy` as held. This is a possible mechanism, not the observed refusal reason. Fresh owning API/route evidence is being captured before changing the guard. No restart or admission override follows from the browser fallback.

A disposable reproduction establishes the HTTP false-hold mechanism with genuine initialized SQLite authority and a real second-thread Writer lock. The unchanged HTTP wrapper observes `writer-busy`, returns the hold page without calling its handler, and returns to READY after release. The bounded correction will allow only that transient busy result into existing Writer admission and retain durable-hold, missing-authority and incomplete-startup refusal. Live route evidence remains separately qualified.

The independently reviewed stage6 constructor fixture executes against the accepted Mylar image with no application mounts or network. Actual ScopedDocker and the real Parent constructor succeed; the canonical 100-leaf installed SDK matches the effective configured admission overlay, and root originals remain unchanged. Owned-container cleanup and original tails complete before the success summary. Independent review verifies 23 artifacts, 78 runtime resources and 41 host controls without skips. This closes the constructor/overlay milestone, not Birth, same-pipe, NFS, archive terminal or application-readiness acceptance.

The corrected read-only DDL observer passes with 97 native rows: 26 Completed, six Failed and 65 NZB handoffs. It retains 65 pack-capture records across generations, which is not a count of completed packs. Its genuine health response reports `writer-busy`. Immutable source/configuration, accepted image/epoch and private snapshot closure pass. The actively changing `ddl-control.json` is explicitly a point-in-time advisory snapshot, with full original hash/identity through its read and later-change reporting. No pack discovery, command-cursor update, scheduling or acknowledgement occurs. Selected-capture and recovery acceptance remain unproved.

The reviewed HTTP/API contention fix passes all 71 host controls without skips, including real Writer/SQLite waits and refusal after pending state, removed authority or incomplete startup. DDL preference recovery passes ten actual-script controls. Complete integration exposes a stale native-source pin in the retained pack worker; the corrected canonical worker, byte-identical fixture and reader-cohort row preserve all function bodies. All 889 manifest rows and 38 worker-fixture pins match. The complete corrected API runner passes 247 controls without skips. Historical precursor and ordinary-observation fixtures retain their original pins. Matching installed-image CI and live route acceptance are still required.

Independent combined source review finds no actionable defect in the HTTP/API admission, preference retry, or paired source-pin closure. Ten actual JavaScript controls pass without skips. Full repository validation and all-file local CI pass, including workflow security and secret scanning. These are source/local acceptance results; published-image CI and live queue-route verification remain pending.

## Queue admission delivery and selected pack preflight 2026-10-13

PR #334 merges normally at `640903a` after all 13 current-head checks pass, including both custom image builds and CodeQL. Independent source review covers the HTTP/API guard, GET retry and two CSS fixes. There are no unresolved review threads. The clean primary checkout is fast-forwarded to the accepted GitHub main. Trusted-main publication and configured deployment remain pending; no new live update or pack import is accepted.

The first pack snapshot summary misses Mylar’s declared uppercase `ID` column after conversion from case-insensitive SQLite rows to dictionaries. The corrected diagnostic normalizes column names and refuses case collisions, with a real declared-schema fixture. It identifies seven completed pack entries and preserves all capture generations. The private metadata-only preflight then passes on the same healthy native image and epoch: all seven latest outer files and ZIP companions exist. Worker maintenance, pack import and auto-import are configured on, with Mylar, Writer and cache mappings present. Native workflow pack policy is not observed by this metadata check. No archive contents are hashed, directory contents scanned, commands fetched, captures discovered or imports acknowledged.

The smallest available outer file is the 26,135,097-byte Rick and Morty Infinity Hour #1–2 capture. This is a candidate for the next exact selected read and generation check, not an operational grant. Source inspection confirms native post-processing defers captured packs to the maintenance worker for inventory. The inactive worker explains why zero-member captures cannot advance. The existing selected-pack pass can operate separately from experimental bulk factories; current configuration, exact source/generation, matching images and affected-state preservation still need proof before execution.

Stage7 implements the Parent/session, finite projection, retained NFS adapter, namespace/epoch handlers and optional installer with 60 host and relocated controls. The corrected finite lease also passes genuine competing-flock expiry without a success ACK. A disposable real-Main fixture is being prepared for application startup evidence. These are source milestones; Birth, NFS, terminal and live library acceptance remain open.

Separate Pull-list PR #335 restores original alias-row ordering after independent review exposes the unchanged validator’s ordering requirement. Its complete 191-control host runner passes without skips using the existing public archive backend selector. Hosted current-head image checks remain required.

The selected Rick and Morty preflight refuses at `authentic-original-selected-packWork` before source hashing. No selected action or import is accepted; a bounded response classification is pending. Independent stage7 review also finds that an early handler/owner closure failure can leave the outer input open before the pump failure flag is set. The source checkpoint is held for a retained-original failure-finally correction, with no new runtime grant.

The unchanged selected guard’s diagnostic successor confirms an API failure with the exact public `Publication authority held; authenticated review required` rejection. This is not a capture-mismatch result. Current hold reason is not established by that rejection; source hashing and import remain unaccepted.

The later unchanged selected preflight succeeds twice against the current daemon. DDL `274610` retains its exact original full row and rowid; the installed source-state function confirms both captured metadata and content generation for the outer archive and companion. Native pack automation is enabled. The private receipt is `teotub47`. This point-in-time result supersedes the current refusal status without explaining the earlier transient rejection, proving row incarnation or accepting an import.

The matching trusted-main worker image is pulled and its exact Linux/amd64 identity is verified. Its 36 installed-module and 689-test isolated gate is prepared; execution and live deployment remain pending. PR #335 fixes an actual CodeQL parity-test regex finding with case-insensitive script handling and a fourth native regression test. Current-head hosted security and image checks remain required.

The actual matching worker gate succeeds against `640903a`: 36 installed source hashes and import origins match, passive factories retain their default restrictions, and all 689 original tests pass without skips. The isolated container is removed by its original ownership-checked cleanup. Receipt `73segxrb` grants no configured-worker or live admission. A separate fresh native baseline attempt stops at a verification guard without lifecycle changes; its diagnosis remains open.

The genuine disposable Mylar fixture also succeeds against the accepted current image. Its actual backup helper verifies isolated restoration before authenticated fresh initialization, then a new ordinary Main starts without an external Writer or database-update bypass. All 175 known canonical source leaves, complete census and native SQLite integrity pass. Result `f0nqb9k2` is an application-startup prerequisite only; Parent/Birth, NFS and archive terminal acceptance remain false. The reviewed supervision correction separately passes all 26 affected source controls and real child-input failure cases.

The fresh baseline diagnostic identifies the unchanged `fresh-native-ready` guard. That predicate combines response success, publication READY/verified and census shape, so the failed assertion alone does not identify which condition failed. Its existing-response projection is awaiting authentication. This is separate from the earlier successful selected pack API/source checks, and no update or lifecycle action occurred. PR #335's later test-only HTMLParser correction replaces the script-tag regex and passes five native UI controls. Current-head security checks and native image verification pass on `25b0138`; the image build/publication remains pending.

The trusted-main `640903a` publication completes successfully for both native and worker images. The native image has not yet been pulled or verified locally. The worker's actual 36-module/689-test result remains isolated proof, with no worker start or configured admission.

Root integrates the 57-path maintained Parent, projection, installer and catalog-caller checkpoint into the owning worktree after verifying all 837 frozen pins, exact tracked preimages and patch applicability. Fresh owning-worktree runs pass all 727 recovery controls and eight preservation controls without skips. `make validate` and full `make ci-local` pass, including the history secret scan with no findings. These checks accept the source checkpoint, not its connected runtime. The trailing blank line is removed through a whitespace-only AST-equivalent correction with installer and manifest repinning; all 13 configured-installer controls and `git diff --check` pass afterward. The catalog caller now uses both `calledfrom='maintenance'` and `pullupd=True` to suppress unrelated search and weekly-list scheduling. Four mandatory controls include the old scheduling failure and the remaining upcoming-issue Wanted behavior. This is not a global status-policy change.

Connected NFS acceptance remains held by a confirmed contract defect: the retained path makes Docker/API calls and starts its probe while the actual native Writer is held. Source origin does not exempt it from the no-remote-call requirement. Separate source owners are correcting lease/probe orchestration and Parent/engine enforcement. Existing source controls do not prove that corrected connected path, so T022/T026/T029 and the dependent live tasks remain unchecked. The selected ordinary-pack recovery path continues separately.

The selected-only pack proposal is also integrated after Root verifies all 83 frozen pins and eight exact worktree preimages. Its eight distinct controls include three complete owning `Packs.member` calls with real matching, SQLite and ZIP inputs. Unmatched or ambiguous catalog identities refuse before catalog/API/import callbacks, including stale ready catalog results. A matched downloaded issue reaches the unchanged import-admission guard. Extraction and landscape inputs remain fixtures, and successful import/ACK is not claimed. Mirrored native fixtures and finite copied maps remain synchronized; the worker image gate explicitly includes these controls. Independent source review is clear. Root's focused selected/pack-only/recovery run completes 52 tests: 35 pass and 17 existing installed-image tests skip on the host. The broader host worker run is interrupted during existing duplicate-archive inventory and is incomplete; it is not a passing 697-test result. Current-image regression verification remains required. No selected pass or live catalog mutation ran.

PR #335 merges normally at `a8e4b61` after all 12 current-head checks pass and both security review threads are resolved. The accepted tree equals the tested head. The primary checkout is clean and fast-forwarded to that revision; the authorized UI owner independently verifies the Gitea mirror at the same revision. Trusted-main image publication has started and remains pending. Live deployment still awaits the existing system-authentication prompt. The source integration branch retains its separate work in progress. Its subsequent local CI run rejects files changed by Root's concurrent task-note edits, despite passing validators; a stable-file rerun is required rather than treating that result as success.

The stable-file local CI rerun passes, including repository validation and the history secret scan with no findings. The pending read-only diagnostic then executes: its READY/verified/nonempty-census predicate passes, but the borrowed rollout version predicate refuses before installed-source and catalog verification. That predicate combines Docker install type, current and latest release equality, and an absent commits-behind value. The refusal alone does not identify the failed subcondition. A fixed-field existing-response classifier is being prepared; no guard is weakened, and no lifecycle or pack action occurs.

The selected preparation follow-up removes the unnecessary `workflowCommands` fetch that previously rotated an unrelated command cursor. The complete owning `Maintenance.prepare` retains health, exact selected `packWork`, reader snapshot and `NativeBatch` checks; malformed or partial selectors refuse before an API attempt. Normal maintenance polling remains unchanged. Root verifies all 84 frozen pins and six preimages before integration. The focused run completes 57 tests: 40 pass and 17 existing installed-image tests skip. Five new prepare controls and the previous eight selected-member controls pass without skips. API responses remain fixtures, so this source result does not prove current live commands, imports or ACKs.

The joined lease/Parent checkpoint is integrated after all 739 pins and 16 current preimages verify. Its initial 755-control gate passes, but independent review reproduces a missing outer-FIFO closure before transport-block removal. The correction checks actual original PID, stream objects and descriptor five-vectors after physical callbacks. Root verifies its 727 frozen pins and ten preimages, then the actual worktree preservation gate passes all eight controls, including the complete copied 759-control run without skips. The retained failure probe now refuses descriptor substitution. Count documentation is corrected to distinguish the 28 earlier executions from four later outer-dialogue cases. Source enforcement and local physical release are accepted at this checkpoint; Parent/Birth, prepared-probe integration, actual NFS and terminal acceptance remain open.

The safe version projection identifies a populated commits-behind counter with Docker/current/latest release predicates all true. The exact expected current API and versioncheck sources confirm that the API returns that counter directly and integer zero is the up-to-date case. The independent read-only observer now accepts only null or exact integer zero while retaining authenticated success and both exact release checks. Eleven observer and four classifier controls pass without skips, with independent source review. Original rollout/helper code is unchanged. The fresh complete baseline check then succeeds: READY, two identical 206-source/Main observations, preserved accepted catalog, unchanged reader and inactive worker. Private receipts `_pyso3sx` and `63wakcx0` establish the current original baseline only, with update and lifecycle flags false. PR #335's trusted-main native image publication also completes successfully; its image pull is awaiting authentication before installed-image and configured route verification.

The actual integrated 759-control source checkpoint also passes full local CI with no optional skips, including the history secret scan. Prepared-probe and active-observation source owners have implemented the remaining call-order split privately; independent standalone source review is clear, while joined registry/installer closure and genuine configured Parent/NFS execution remain required.

The published `a8e4b61` repair image pull completes successfully, digest `de011571…`; its exact local identity check is pending authentication. Root verifies all 842 proof/preparation artifact pins and 1,032 resource hashes, and independently runs all 13 unchanged preparation controls without skips. The narrow rollout-helper successor accepts null or exact integer zero while preserving authenticated Docker and both release checks, backup, rollback and lifecycle functions. Root verifies all 32 pins and runs 28 current plus one retained historical control without skips. Target-image helper verification and fresh bound rollout remain required.

The prepared-probe/phased source batch passes its author’s 799 recovery and eight preservation controls. Root verifies all 757 pins and 24 owning preimages and reviews the production delta, but does not integrate it: independent review finds that the post-release original-path check still requires canary directories legitimately removed before release. The frozen proposal remains held while a finite original-bound absence transition is implemented. This is a retained source failure, not connected Parent/NFS acceptance; the owning source remains at the accepted 759-control checkpoint.

The finite V9 operational dependency joins preserve the original backup/restore, selectors, Writer release, startup and rollback algorithms. Root verifies all 27 pins, reviews the literal-only production patch, and runs 46 driver plus 19 capture/preparation controls without skips. The first direct isolated-Python driver invocation fails before collection because `-I` excludes its sibling test imports; the corrected runner explicitly adds only that frozen public test directory, with all assertions unchanged. The latest owning-source full local CI also passes, including a 601-commit secret scan with no findings. These are host/source results; target-runtime checks, actual image identity and fresh operational bindings are still pending.

Independent delivery review clears the applied 75-path checkpoint and verifies the source maps, selected canonical/fixture parity and unchanged 55/72 task markers. The normal commit hook then refuses five credential-shaped public source-hash rows in the base-source JSON; no commit or push occurs. A semantic-equivalent serialization correction is being prepared without scanner exclusions. The prior unstaged local CI result does not substitute for this staged check.

The source-map serialization correction passes independent review and Root’s exact parsed-data comparison, preserving the historical `f250ebd6…` preimage and strict successor byte pin. All 13 owning installer controls pass. The corrected phased/cleanup batch is integrated after Root verifies 762 pins, all owning preimages or the three explicitly coalesced serializer successors, and every final source hash. Independent 20-control review is clear. The owning preservation run passes eight controls, including all 804 recovery executions without skips. A trailing blank line remains to be corrected with its hash references before commit. Genuine configured Parent/NFS acceptance remains open.

The published repair image identity is now observed as `de011571…`, Linux/amd64. Its inherited Docker revision label names upstream base source; acceptance uses the actual installed-byte and runtime checks, not that label. The image-bound verifier changes only its previously disabled candidate-image assignment; the combined source/cohort, UI, import and recovery batch is running in disposable containers without live application mounts. No service update, import or ACK is accepted by this checkpoint.

The first EOF/hash successor passes its focused peer checks but fails Root’s complete owning gate: the active caller retains the old lease-interface digest. The unchanged factory-entry assertion identifies that exact stale reference. The peer withdraws its full-clear claim; the limited AST/map evidence and failed run are retained while the full consumer graph is corrected. No commit or source delivery is accepted. Separately, the actual image batch passes the preceding source/cohort/helper/rollback/startup stages, then exits at `api247`. Its retained terminal-family excerpt shows callbacks not reached, so the original first error is being read before changing code or fixtures. Mylar and Komga remain unchanged.

The API failure’s first original error is `KeyError: ARCHIVING_UTILS_ROOT` in the retained composed-worker fixture. The first 42-test family passes. Mylar’s checked Docker build supplies the fixed installed utility path, but the final-image verification command did not supply that test environment setting. Its absence prevents the later retained callbacks from being reached. A bounded source audit and fixture-profile correction is underway; all 247 assertions remain required, and no runtime-image success or deployment is claimed.

The complete interface-hash correction passes Root’s 722-pin and four-preimage verification, then the owning 804-recovery/eight-preservation gate without skips. The normal staged secret scan also passes on all 84 owned paths without exclusions; this includes the newly changed hardlink-child source as well as the added fixtures. The failed narrow hash correction remains historical evidence. Final stable CI, source commit and latest-main reconciliation are still pending.

The corrected published `a8e4b61` image completes all 12 isolated stages. Receipt `14ah40ov` binds actual 206-source parity, all 247 API and five Pull-list controls, helper/rollback/startup checks and 17 genuine import/recovery cases to image `de011571…`. The earlier missing test-environment failure is retained; no assertion changes were made. This does not claim deployed UI behavior or live pack recovery.

The 84-path source checkpoint commits as `7aac7e0` and reconciles current main as `d0752af`. Two hash-map conflicts are derived from the merged files; all 396 rows in each map match source and both UI/catalog gates remain. Full local CI passes. Independent behavior and security audits find no code defects; two stale contract paragraphs are corrected. Publication scanning covers 3,055 tracked files without secrets or skips, with no privacy indicators in changed files; Gitleaks and exact-range TruffleHog also pass. PR [#336](https://github.com/bolens/homelab/pull/336) starts normal hosted checks. Source lint identifies two unused test imports; their removal and manifest repinning preserve assertions and production code, with affected controls and the complete owning 804/eight gate passing. Current-head hosted acceptance is required before merge.

Hosted native-image checks expose two additional fixture-delivery defects: three stale entries in the unchanged full worker hash-map assertion, and an absolute private predecessor path absent on CI. The correction repins the three actual worker hashes and dependent finite maps, and packages the byte-exact reviewed predecessor with a specific Docker-context exception. The parity test changes only its locator; all assertions remain. Root verifies 3,076 artifact pins, seven exact preimages, the historical Git blob and every one of 38 worker hashes before applying it. Earlier no-finding delivery review is qualified by these discovered seams; fresh-head hosted native/API checks remain mandatory.

## Reviewed source delivery 2026-10-14

PR #336 merged through the normal squash workflow as `dcd57047dd0bff12f624d405e550e5a6a4bc9ab8`. Its tree matches reviewed and tested head `5968d8426764835da6578948a5a803c9245edb9d`. All 14 hosted checks pass, including both native and worker builds, complete image gates, security and source lint. Final review-thread and open CodeQL-alert queries return zero. GitHub, local main and the authorized Gitea mirror share the merge; no open repository PR remains at this observation. Trusted-main matching-image publication is running and has not yet been accepted.

The corrected private rollout helper and literal joins pass 107 executions in the actual Python 3.10 image, with no skips: 29 helper controls, 65 join controls and 13 preparation controls. Some controls overlap; these totals are executions, not distinct new behaviors. Earlier failed runs remain retained. This establishes runtime compatibility of the helper/proof inputs, not new-image or configured recovery acceptance.

A fresh independent read-only baseline confirms ordinary Mylar READY, preserved catalog, unchanged Komga and an inactive experimental worker. No service lifecycle or application-data mutation occurred. T022/T026–T029, archive continuity/terminal acceptance and selected live pack ACK remain open. Their full acceptance conditions are not satisfied by source merge, isolated controls or a healthy baseline.

## Configured-base source correction 2026-10-14

Independent review of the merged-image preparation finds the optional configured reader installer still expecting predecessor API, native-writer and pack-catalog hashes. Its unchanged strict SDK join therefore refuses the current source cohort. The finite correction derives those three expected installed hashes and two cohort provenance rows from the maintained patchers/current sources, repins the dependent checked manifests, and retains the historical preimage. All installer guard functions and the original thirteen tests remain unchanged. Three added controls prove the current complete join, refusal of each predecessor digest, and preserved source/completeness/default-acceptance guards.

Independent review verifies all 738 proposal pins, nine exact preimages, 175 known sources, 100 required SDK names and complete dependent maps. Sixteen installer controls pass independently; Root repeats the owning eight-control gate, including all 807 recovery executions, without skips. No configured or installed Parent acceptance is asserted, and image/operational bindings remain disabled. The ordinary published-image proof remains separate.

PR #337’s task-documentation update was merged while two validation jobs were still running; this did not follow the required wait-all policy. Both subsequently pass, and all applicable hosted checks are now complete. GitHub, local main and Gitea share its reviewed tree. The isolated worker wrapper failure is recorded: all 702 application tests and 13 selected controls pass, but buffered output places the earlier summary last and correctly fails the terminal-summary guard. A two-flush correction preserves that guard and passes the repeated 702-case suite, thirteen repeated selected controls and all 36 installed-source/import checks without skips. The live worker remains inactive; configured acceptance is false.

## Host reboot checkpoint 2026-10-14

The host rebooted before temporary evidence preservation completed. The repository changes survive, but temporary proof packages, private helper source, logs and peer receipts were cleared. Prior execution results remain historical reports in this evidence history and the conversation; their raw temporary artifacts are unavailable for reinspection. No configured or live acceptance is inferred from them. The matching PR #336 native and worker publication completes successfully on GitHub.

The surviving source patch and resume notes are now saved on private persistent storage. Further operational helpers and evidence must use persistent storage; the project guide records this correction. Reconstruct missing packages and recapture current service/process/source and file bindings before any rollout or selected pack recovery. No new image was deployed or worker activated before the pause.

The fresh post-reboot Docker inspection finds the prior Mylar image running but unhealthy; its saved health probes report a publication-authority hold. Komga is healthy on its prior image, and the experimental worker remains Created with no process. The hold cause is under read-only diagnosis; this does not establish startup completion or post-reboot catalog preservation. A fresh source audit of PR #338 verifies all checked maps and sixteen installer controls without skips. The exact historical API predecessor bytes remain unavailable, so the committed API expectation is not relabeled as a newly reproduced derivation.

## Post-reboot authority diagnosis 2026-10-14

Fresh read-only health and installed-source tracing identify `authority-unavailable`: `initialization_witness` rejects the immutable bootstrap binding. The exact comparison finds changed device numbers for the database, writer directory and lock, with matching inodes, schema, epoch and bootstrap token. These observations cover the examined witness, not every subsequent registration, marker or tagger binding. Existing source checks each committed registration and the final marker independently; restarting alone does not repair this mismatch.

The diagnostic changes no authority records, pending markers, databases or services. Source review finds no maintained device-only adoption route. Recovery must preserve immutable registration history and verify affected-state backup/restore and filesystem provenance before any identity transition. Separate tagger identity checks and current catalog preservation remain unverified. T017 and operational acceptance stay open. Private diagnostic source and concise results are retained on persistent storage.

The reviewed-reconciliation contract permits an explicit new destination trust decision supported by a complete current backup and isolated restore, while acknowledging absent historical filesystem identifiers. A frozen source proposal preserves all immutable SQLite rows, marker and tagger identity bytes through a bounded append-only device-transition record. Production readers, crash controls and operational backup/restore remain unimplemented or unexecuted; the proposal grants no admission. Independent native-verifier and selected-observer source reviews pass, but their image and configured executions remain pending.

The corrected existing-only tagger diagnostic executes successfully in the current native container. Its four directories share the current database device, and substituting the immutable bootstrap's original device number reproduces the unchanged stored identity digest exactly. The current digest differs. All six examined normalizer/tagger/release/publication/derivative pending-marker paths are absent. This proves the separate tagger tuple mismatch, not journal/staging terminality or migration admission. No state is changed.

The published PR #336 native image is pulled and its exact linux/amd64 identity is checked. Reconstructed source capture verifies the 175-source expected base and three public authentication leaves, and the installed cohort/type check passes. The combined reliability stage then fails the unchanged Mega plaintext-allocation assertion. Source tracing confirms `patch_cache_admission.py` changes the library outside the native package, while the final Docker stage omits that checked library. This is a final-image packaging defect, not successful full-image acceptance. A corrective copy-boundary audit and regression are underway. The failed run and logs are retained privately; no deployment occurred.

The first private device resolver batch passes seventeen host controls in independent repetition, including genuine SQLite history and future registrations. Peer review subsequently finds absent-ledger input drift; that version remains rejected and a guarded successor is being tested. Neither draft is installed or used for authority admission.

The hosted PR #338 build fails a separate scheduling defect: the outer 90-second reader-control deadline is shorter than one nested child's existing 120-second allowance and kills the unchanged 807-control aggregate. The corrective runner retains ordinary 90-second and child 120-second bounds, giving only the two aggregate paths a finite 6,780-second cap derived from 56 serial children plus fixture overhead. Overall workflow and isolated-launcher limits remain independently tighter. Ten owning controls, including all 807 checks, pass without skips; independent review preserves original assertions, suites and actor policy. Leaf-first repins change only two standalone rows and their three API-map references.

The final-image correction copies the actual checked Mega library into the runtime stage and checks its allocation, temporary-file, before-write and promotion calls there. The mandatory CI regressions replay the owning patch and final COPY routes: the original image contract fails, while fifteen corrected controls pass with no skips. Independent source review confirms only the Mega library was omitted among the audited external patch outputs; existing template copies and Main permissions stay intact. Corrected source still requires full local checks, hosted CI and a fresh built-image gate before deployment.


## Recovery delivery checkpoint 2026-10-14

PR #338 is squash-merged as `454b41df` after every applicable hosted check
completes successfully. Full local CI passes. GitHub, local main and the
configured Gitea mirror share that revision. The earlier scheduling failure and
missing final-stage Mega library remain recorded above. Corrected trusted-main
image publication completes successfully. No live update is claimed.

The current Mylar container successfully repeats the read-only Btrfs identity
query as UID 1000. Its stable filesystem/subvolume result matches the host
observation. Identifiers stay private. This proves current query capability,
not recovery authorization, historical continuity or actual remount behavior.

The native recovery batch is integrated locally after independent source review.
Root repeats all 57 controls in both the frozen package and owning checkout:
22 new controls and 35 unchanged bootstrap/registration controls pass without
skips. A retained peer failure is corrected: mutation of the original accepted
plan now refuses before fence removal and preserves the fence, SQL and marker.
Tests use real SQLite, Writer locks and future registration/restart. Portable
kernel-byte fixtures supply no real filesystem or remount acceptance.

The passive worker source passes five separate-process controls plus a full
registry/current-marker route before and after genuine registration. Its reader
has no native maintenance writer or held-journal allowance. Worker packaging and the latest native-source join now pass source verification: regeneration reproduces all twelve outputs, with four additional separate-process stable-branch controls. Matching worker-image verification remains pending. The reconstructed main-image verifier
also passes independent source review and four host controls with image binding
unset. CI delivery, installed-image gates, verified backup/restore and live
recovery remain open under T017. No aggregate task is checked off on fixture
results alone.


The integrated packaging closure retains all