fix: Report unsupported scheduled preference maintenance

[buildingjoshbetter]

Scheduled maintenance and bulk ingestion reported elapsed success after calling a preference helper that immediately returned an empty result. Report UNAVAILABLE (ScheduledPreferencesUnsupported) for that unsupported stage and classify otherwise complete maintenance as completed_with_limits. Refs #746; stacked on #828.

Changes

Real layer/style failures, deferrals and pending caller writes retain precedence. Captured cancellation stays cancelled after its event clears. Preserve the nine full-maintenance result keys, no-attempt/style-only behavior, missing-personality handling during ingest and direct entity/query-time preference retrieval. No preference-generation persistence, schema, model or retrieval change is introduced. Personality changes are documentation only.

Prepared publication head: bf0f38c4cac55c8ccff1d39b19eacb9fe9213bbf. Proposed remote refresh destination for draft #829: fix/preferences-status-publish-v1. At preparation, the current remote #829 head remains 39a8827ac01635652fdb2416896c7fac79f4bf73; the proposed new head is distinct. Intended base: draft #828 through fix/tier-public-publish-v1, exact prepared Unit C head 5873e5a4978f37bcd31772d6840717fb217df713.

The feature delta from that Unit C base is nine paths: three production, five tests and one documentation file. The #746 status behavior remains unchanged from the reviewed earlier preference implementation; reconnect and MCP fixture changes are inherited from the refreshed base.

Preference controls and assertion preservation

The original 11 stdlib/in-memory preference regressions and 47 independent AST/in-memory controls pass. Fixture corrections bind the actual production constant and replace obsolete no-op-success expectations with the exact unavailable status and zero helper calls. Existing methods and meaningful assertions are preserved.

The routing fixtures retain 86 test definitions. They preserve 456 assertion calls and replace two obsolete checks with eight checks for the eight pending layers, exact unsupported ninth result, all nine keys and zero preference-helper calls: 456 + 8 = 464 current assertion calls. This is a source AST inventory, not a count of runtime assertion executions. Original failed source and logs remain preserved.

Inherited reconnect and MCP correction

The refreshed base adds the scoped #684 closed/I/O cached-handle admission correction: exact Engine and failed-handle proof, strict automatic-owner receipt validation, handle identity rechecks before/after waiting and one shared 10,000 ms budget. Current-owned-writer refusal occurs before opening, avoiding a nonreentrant self-acquire. Manual/receiptless owners, serving/already-admitted callers, replacement handles and owned writers retain refusal. Strict maintenance admission source remains unchanged.

The historical macOS closed-handle failure and old #829 Windows I/O failure, job 114395231281, reached maintenance-owner refusal. Their actual owner identities and receipts remain unproven. Duplicate hosted jobs passed. Separate synthetic actual-function controls reproduce the verified automatic-owner condition; they do not establish the historical CI actors. Original macOS and Windows logs remain preserved.

The shared reconnect controls pass 26 methods and 42 independent controls. Three selected regression assertions observe GC enabled at their checkpoints. All 101 boundary methods and 550 unittest assertion and fail call AST nodes remain unchanged through collaborator repairs; the complete memory-only boundary run passes 101/101. Original 21 missing-import errors and nine inherited setup errors remain preserved. Actual production exception/budget bindings and a strict unexpected-owner-wait guard are retained.

Original combined full source 5cd23db000fd2a7c98e93395be33cf8638641db1 failed ten MCP fixture tests while 3,725 tests and 3,223 subtests passed. Nine failures were missing time in the actual opener AST namespace; one parallel-wait assertion was the resulting cascade. Original source, container and full log remain preserved, log SHA256 2a706b32240afa808f582e1751beeea1a24302f1a6f86f3eb3eddec5fa85befd.

The final MCP-only fixture repair binds actual stdlib time, actual writer-lock/maintenance-error class ASTs, the actual storage budget and explicit inert imports with strict unexpected-wait/import guards. All 11 test methods and 52 unittest/Mock assertion and fail call AST nodes remain unchanged. The safe suite passes 11/11 without SQLite connects and independently frozen MCP controls pass 35/35. Production, model, retrieval and GC behavior are unchanged by this fixture repair.

Exact-source validation

Gate	Exact source	Passing tests / subtests	Runner elapsed	Full-log SHA256
Expanded focused	3a9360b1ba6bf210a1b54e05287117923fbbaae5	833 / 633	39.36 s	e1c203f112a3dedc0411874785f00a528c6bd3d34809657f1903df46da7ee0df
Combined full	3a9360b1ba6bf210a1b54e05287117923fbbaae5	3,735 / 3,223	412.643 s	a750f39645d1c866e5e68cddaa9ad49fe3616f3500be6180dadbbb4344dbeaed
Exact #746 full	bf0f38c4cac55c8ccff1d39b19eacb9fe9213bbf	3,542 / 2,424	411.058 s	065e27e9138fd77c458aae6ee6af88b5032ca6697363194b7e29c01dcbd8796b

Both full gates exited 0 with 17 skips and three network deselections, no timeout/OOM and matching actual source/build/log hashes. They ran sequentially in offline installed-wheel images with eight CPUs, 16 GiB, read-only source/models/fixtures and enabled native SQLite flags. No GC-disabling override was used; three selected regression assertions observed GC enabled at their checkpoints. Full-run runtime GC state was not recorded. All-test installed import origin remains unmeasured.

Exact publication source/build/run: source-preferences-status-publish-v5, build-preferences-status-publish-v5, preferences-status-publication-v3. Build log SHA256 6338cb64f7833086e8a4d4034d935e361cb4fa7374c70486ac3b45fbdb7db138. Receipts: .work/reports/preferences-status-publication-validation-v3.json, .work/reports/preferences-status-publication-container-config-v3.json, .work/reports/build-preferences-status-publish-v5-evidence-v1.json, and current combined/focused receipts.

New-head hosted CI remains pending. Validated private integration remains 7bd70e9fa4812e9925b86652dbe5c3a5c92bafe1; adoption of the prepared combination is separate. Persisted cited preferences, native host/model-visible acceptance and the original 26.24 GB Mac memory fix remain unproven. Complete-archive Deep Search and durable source citations remain future scope. Zero accepted issues are closed; broader #746 acceptance remains open. Draft only; no merge or release.