hygiene(tick-history): 2026-05-01T22:24Z drain tick + 1916Z orphan backfill

docs/hygiene-history/ticks/2026/05/01/1916Z.md

@@ -0,0 +1 @@
+| 2026-05-01T19:16:00Z | opus-4-7 / autonomous-loop tick | 98fc7424 | Pipe-escaping cleanup tick. **PR #1130 MERGED at 19:13:47Z** — detect-changes pattern memo (sibling-repo parallel-optimization external anchor) now canonical. **#1141 (1907Z tick-history)** real markdownlint failure — pipe characters inside backtick code spans broke GFM table parsing. Earlier "false-positive" classification on #1136/#1138 was correct as far as it went (the leading-prefix claim was wrong) but I missed a separate-class issue that #1141 surfaced explicitly: literal backtick code spans like backtick-1-pipe-backtick or backtick-pipe-2026-backtick inside a body cell DO need backslash-escaping (the actual pipe character must be backslash-escaped even inside code spans for GFM table compatibility). Fixed by replacing pipe with backslash-pipe in code spans + removing trailing-space code-span. **#1140 (1903Z tick-history)** same pipe-escape issue in `xargs` example. Fixed. **#1125 (B-0154)** xref to `feedback_git_native_vs_github_native_*` was missing `memory/` prefix. Added. Cron 98fc7424 healthy. | [#1130 MERGED via auto-merge; #1141 + #1140 pipe-escape fixes pushed; #1125 memory-prefix fix pushed; 4 threads resolved] | The pipe-in-code-span class is now-clear: GFM table cells parse `\|` literally only if backslash-escaped, even when the `\|` is inside backticks. Earlier shards may have the same latent issue (#1136 / #1138 unmerged); will surface on next round of Copilot review. Mitigation pattern: when authoring tick-history shards, pre-escape any pipe inside backticks. Lint catches it (markdownlint MD038/MD-table). Composes with the prefer-mechanical-anchors discipline — Copilot reviewer caught on the *new* PR what I had categorically dismissed on the *prior* PRs. The lesson: same-shaped finding may have a different actual class than my first-look classification. Verify-before-categorizing IS the discipline; my "false-positive" tag on the earlier PRs was premature dismissal. |

docs/hygiene-history/ticks/2026/05/01/2224Z.md

@@ -0,0 +1 @@
+| 2026-05-01T22:24:00Z | opus-4-7 / autonomous-loop tick | 98fc7424 | Drain-focused tick: post-compaction wake. **PR #1173 (B-0159 backlog row) MERGED** during tick — refresh-github-worldview / repo-state two-layer architecture is now on main. Two Codex P2 fixes shipped before merge: `--since` placement before pathspec separator, and output-contract reconciliation (single-JSON canonical with `summary` field, supersedes earlier two-pass framing per Deepseek). 5 threads resolved on #1173 (2 Codex P2 in-commit, 2 Copilot dangling-pointer cleared by #1171 main-merge, 1 Copilot output-contract). Then drained 2 simple tick-history shards: **#1103 (1111Z) + #1104 (1115Z)** — Copilot caught "force-with-leased" used as past-tense-verb when `force-with-lease` is the git flag-name; fixed to "force-pushed with lease" in both shards, auto-merge armed. Sweep across 17 sibling tick-history branches for the same typo class returned false-alarm: all matches pointed to 1346Z.md on main, where "force-with-leased grammar typos" appears as a *meta-reference* about the typo class, not as the typo itself — pattern-match isn't a substitute for per-hit verification. /btw aside on "where is the eat document?" answered (EAT = `docs/research/economic-agency-threshold-2026-04-27.md`) then re-classified by Aaron as personal-register Claude.ai concern, not factory work. Cron 98fc7424 healthy. Backfilling orphan 1916Z shard (pipe-escape fix recap from prior wake) in the same PR. | [#1173 MERGED via auto-merge after CI green; #1103/#1104 armed; 7 threads resolved across 3 PRs; force-with-leased typo fixed in 1111Z.md + 1115Z.md] | Cites BLOCKED-with-green-CI rule (CLAUDE.md) — `bun tools/github/poll-pr-gate.ts <PR>` reporting BLOCKED + 0 failed required + armed auto-merge means *go look at threads first*; #1173 thread-list query took 2 seconds, dispatch took 30 seconds. Cites refresh-before-decide invariant (PR #1171) — multi-PR refresh via `poll-pr-gate-batch.ts --all-open --summary-only` surfaced 27-PR queue (19 thread-blocked) before tick-selection. Cites pattern-search-isn't-pattern-match: 17 branch hits collapsed to 1 file on main with meta-reference; per-hit verify-before-fix prevented spurious churn. Substrate-density observation: 27 open PRs is significant — drain-rate vs gen-rate inversion noted, no new substrate generated this tick beyond required fixes + this row. |