[Backend][Proposals] Expiry-sweep transition dedupe lives on the DI-scoped AutomationProposalService and STATUS overstates it (from PR #2505 review)

[Chris0Jeky]

From the fresh-context review of PR #2505 (#2209), non-blocking. (1) MEDIUM: AutomationProposalService._lastSkippedArchivedBoardCount (~:46, used ~:1247) implements only-on-transition logging, but the service is registered AddScoped (ApplicationServiceRegistration.cs:56, McpApplicationServiceRegistration.cs:33-35), so the field resets per scope; a future per-request caller of ExpireProposalsAsync would log the Information line on every call (the flood the change meant to prevent) and the Debug 'still skipping' branch is unreachable. Latent today: ExpireProposalsAsync has no production caller (worker/controller/CLI/MCP all absent). Options: drop the field from the service (the worker is the operator-reporting lane) or document the lifetime caveat like ProposalHousekeepingWorker.cs:14-19 does. (2) MEDIUM: docs/STATUS.md #2197 bullet says 'both expiry paths emit the non-secret skip count only on transitions' — true for the singleton worker, only per-instance for the scoped service; qualify it, and the inherited test counts in that sentence no longer match the head. (3) LOW: the service's zero-transition sentinel branch (AutomationProposalService.cs ~1257-1261) has no service-level test; ProposalHousekeepingWorkerTests 'nothing withheld' test lost the expirable-proposal scenario and now pins BeEmpty over all log lines; AutomationProposalRepository.cs ~324-325 comment lost the two-reads-can-disagree rationale. Refs #2209, #2197.

[Chris0Jeky]

Claiming bounded lane alpha-product-trust / TD-2520-zero-transition from base 9cf28bbc60925344c77a8cdaed9ad8cc37064634. Scope: add the missing service-level regression for the nonzero-to-zero archived-board skip transition, with a real expirable fixture and precise log assertions. Owned path: backend/tests/Taskdeck.Application.Tests/Services/AutomationProposalServiceEdgeCaseTests.cs only. No DI lifetime redesign, worker changes, repository changes, or STATUS edits.

[Chris0Jeky]

Released bounded slice TD-2520-zero-transition in PR #2637: #2637

• commit 8fc32553074d0746e7c3aaea958a5af07ecda152
• added the service-level nonzero-to-zero archived-board skip regression
• retained a real expirable proposal and asserted expiry, save, notification, precise informational recovery log, and no stale debug log
• final edge-case class: 16 passed, 0 skipped; git diff --check passed
• two coordinator review passes found no blocking findings
• no DI lifetime, worker, repository, or STATUS changes; parent remains open for other rows

Automatic Codex review is unavailable because the account review quota is exhausted. No merge performed.

[Chris0Jeky]

Coordinator note (2026-09-05, D-12 sweep): the Codex session's PR on this issue was reviewed by one fresh-context Opus 5 reviewer (record on the PR) and merged with a merge commit. The PR body used Refs, so this issue stays open; what merged and what remains, from the review:

• PR #2637 merged as 30d715611: application-service regression for the archived-board skip count recovering from nonzero to zero (the one log branch no existing test reached). The DI-lifetime, worker, repository and STATUS decisions this issue carries are untouched.

[Chris0Jeky]

Coordinator disposition (2026-09-06, Codex-session PR carried under the standing D-12 route). PR #2725 (docs: explain separate expiry reads, comment-only in AutomationProposalRepository.cs) merged as 803eba98b, review SHIP. Remaining from the review (LOW): the wording "never inferred from two reads that could disagree" can read as a cross-read consistency guarantee while expirable and skippedArchivedBoardCount are two separate untransacted queries; suggested wording on the PR.

[Chris0Jeky]

Coordinator disposition (2026-09-06, Codex-session PR carried under the standing D-12 route). PR #2745 (Prove expiry transition state is scoped) merged as e6c02dd6f, review SHIP. LOWs: ExpireProposalsAsync has no production caller (the shipped sweep is ProposalHousekeepingWorker, which keeps its own transition state), so the log-noise guarantee does not reach the worker; near-duplicate of the edge-case-file test; no Times.Exactly(3) verify.