hygiene(tick-history): 2026-05-01T20:42Z — review-thread-drain tick

docs/hygiene-history/ticks/2026/05/01/2042Z.md

@@ -0,0 +1 @@
+| 2026-05-01T20:42:00Z | opus-4-7 / autonomous-loop tick | 98fc7424 | Review-thread-drain tick across PRs #1153 + #1155. **REFRESH WORLD MODEL**: poll-pr-gate-batch on the 4 session PRs returned #1154 + #1156 MERGED, #1153 BLOCKED resolve-threads, #1155 BLOCKED wait-ci with non-required dup-link warning. Ran GraphQL queries to enumerate unresolved threads — 6 on #1153, 3 on #1155. Per-class taxonomy: 1 outdated (Codex P2 paginate, already fixed by 275458e), 8 real-fix. **Three commits this tick**: (1) `6b8c743` on #1153 — 4 fixes: requireValue rejects `-` prefix not just `--` (Copilot — `--owner -h` was silently consuming `-h` as value); spawn() drops explicit stdio so child.stdout/stderr non-null typed (Copilot — strict tsc broke); listOpenPRs uses `gh api --paginate ... --jq '.[].number'` for line-oriented primitive output instead of line-split JSON parsing of multi-line page bodies (Codex P2 + Copilot P1 dup); de-name "Aaron 2026-05-01" → "the human maintainer 2026-05-01" in CLAUDE.md bullet + script header (Copilot — persona-name-on-current-state-surface rule, lineage stays in memory/ files). (2) `9e12f3f` on #1153 — Copilot P0: pollAllBounded must never reject. Wrapped `await pollFn(...)` in try/catch, converts rejections to PollOutcome.error with exitCode -1. New test asserts the invariant (synthetic Promise.reject becomes PollError; other PRs land normally). 11/0/131ms. (3) `cea6017` on #1155 — Codex P1 + Copilot P0 dup-link MEMORY.md fix (cherry-pick artifact from cross-branch merge produced two rows for the dynamic-bash file; removed older shorter row, kept newer with worked-example footer); Copilot abs-path fix (replaced `/Users/.../SQLSharp/...` in caused_by with `../SQLSharp/...` per cross-platform-reproducibility rule). All 11 unresolved threads resolved across both PRs after each fix landed. Cron 98fc7424 healthy. | [PR #1153 +2 commits, all 8 threads resolved (1 outdated, 7 real-fix); PR #1155 +1 commit, all 3 threads resolved; both PRs now wait-ci only with no warnings; auto-merge armed; final state confirmed via poll-pr-gate-batch] | The review-thread-drain demonstrated the new tools' value end-to-end this tick: poll-pr-gate-batch surfaced the BLOCKED states; structured `unresolvedThreads` count drove the GraphQL queries; per-class taxonomy (outdated vs real-fix) classified each finding before action; the new TS form factor caught its own pattern-violations (the persona-name finding by Copilot is exactly the kind of issue the SQLSharp `pattern memory file` warns about — js-in-ts shape leaks into other dimensions like attribution-on-current-state-surfaces). The DST tests held across the 4-change refactor + the new throw-recovery invariant — proof that the test suite isn't ornamental. The new try/catch invariant also generalizes: an orchestrator's "must-never-throw" contract is testable in TS, was previously only documentation in bash equivalents. Carved blade reinforced: *"DST coverage isn't a nice-to-have — it's the only way the orchestrator-must-never-throw contract gets enforced."* |