shard(tick): 2026-05-16T03:22Z — PR #3690 markdownlint MD038 unblock + sweep

docs/hygiene-history/ticks/2026/05/16/0322Z.md

@@ -0,0 +1,93 @@
+# Tick 2026-05-16T03:22Z — Otto-CLI
+
+**Surface**: Otto-CLI (Claude Code, Opus 4.7 1M context, autonomous-loop tick)
+**Parent tick**: [2026-05-16T03:16Z](0316Z.md) — PR #3699 round-3 quality
+
+## What landed this tick
+
+PR #3701 (0310Z shard) MERGED at 03:16:50Z. Diagnosed and fixed the long-standing PR #3690 merge block — a **required markdownlint check failure** that had been preventing the PR from auto-merging since tick 6 (~70 min ago).
+
+| PR | Action | Result |
+|----|--------|--------|
+| ~~[#3701](https://github.com/Lucent-Financial-Group/Zeta/pull/3701)~~ | (carry-over) | MERGED 03:16:50Z |
+| [#3690](https://github.com/Lucent-Financial-Group/Zeta/pull/3690) | required check failure diagnosed → fixup `81b2f1a` pushed | OPEN, armed, awaiting CI re-run |
+| [#3699](https://github.com/Lucent-Financial-Group/Zeta/pull/3699) | (no new findings) | OPEN, armed, awaiting CI |
+| [#3703](https://github.com/Lucent-Financial-Group/Zeta/pull/3703) | (no new findings) | OPEN, armed, awaiting CI |
+
+## Why PR #3690 was stuck
+
+Across 6 ticks I'd treated PR #3690 as "armed, awaiting CI" without investigating its check rollup. Today's `gh pr view 3690 --json statusCheckRollup` revealed:
+
+| Check | Conclusion | Required? |
+|-------|-----------|-----------|
+| `lint (markdownlint)` | **FAILURE** | **YES** |
+| `lint (backlog ID uniqueness)` | FAILURE | no |
+
+The required markdownlint check was the actual blocker. `poll-pr-gate.ts` reported `requiredChecks.failed: 1` + `nextAction: "fix-failed-checks"` — clear signal I'd missed earlier.
+
+Substrate-honest takeaway: **don't assume "armed + awaiting CI" means "will merge"**. Required-check failures are a distinct block class from unresolved-threads + non-required-check warnings. The next-action signal in `poll-pr-gate.ts` is the canonical disambiguation surface.
+
+## The MD038 violation
+
+```text
+docs/hygiene-history/ticks/2026/05/16/0240Z.md:20:18
+  error MD038/no-space-in-code Spaces inside code span elements
+  [Context: "`git `"]
+```
+
+Line 20 of my tick-6 shard read:
+
+```text
+Fix: prepend `git ` to both snippets. Diff: +2/-2.
+```
+
+The trailing space inside the backticks (intentional: meant to convey a literal "git" prefix followed by a space) trips MD038. Markdownlint considers a code span with leading or trailing space inside the backticks malformed.
+
+### Fix
+
+Rewrite to drop the inside-backticks space while preserving semantic:
+
+```text
+Fix: prepend the `git` prefix to both snippets. Diff: +2/-2.
+```
+
+Same meaning; backticks now contain only the literal text `git`. Local `bun x markdownlint-cli2 docs/hygiene-history/ticks/2026/05/16/0240Z.md` → exit 0.
+
+## Sweep of recent shards for MD038
+
+`bun x markdownlint-cli2 docs/hygiene-history/ticks/2026/05/16/*.md` on `origin/main` (20 shards) → exit 0. Only 0240Z had the violation. The two in-flight shards (#3703 from tick 12, #3699 the audit PR isn't markdown) are also clean per direct inspection.
+
+## Operational notes
+
+### Required-check vs non-required-check distinction matters
+
+Across this session I've been treating most CI signals uniformly. The required-vs-non-required split is a first-class merge-gate concern:
+
+- **Required check fails** → auto-merge blocks; PR stays open indefinitely
+- **Non-required check fails** → warning in `poll-pr-gate.ts`'s output; auto-merge fires anyway
+- **Unresolved threads** → blocks if `required_conversation_resolution` is on (it is); else passes
+
+The `nextAction` field in `poll-pr-gate.ts`'s JSON output disambiguates these. From [`.claude/rules/blocked-green-ci-investigate-threads.md`](https://github.com/Lucent-Financial-Group/Zeta/blob/main/.claude/rules/blocked-green-ci-investigate-threads.md): when `gate: "BLOCKED"` + `requiredChecks.failed == 0`, the block is unresolved threads. When `requiredChecks.failed > 0`, the block is the failing check itself.
+
+Adding this distinction to the per-tick refresh discipline would have caught PR #3690 at tick 7. Worth carrying forward.
+
+### MD038 is a non-obvious markdown lint class
+
+The inside-backticks-trailing-space pattern is easy to miss — the trailing space is visually invisible in the rendered output but semantically present in the source. Other MD038-triggering patterns:
+
+- `` ` foo` `` (leading space)
+- `` `foo ` `` (trailing space, this case)
+- `` ` foo ` `` (both)
+
+Adding a `MD038` self-check before commit (alongside the MD032 paragraph-before-bullet check) would catch the class. Same pattern as the relative-path audit's pre-write checks.
+
+### Holding-discipline state
+
+Concrete substrate landed (PR #3690 unblock via fixup). Counter resets per "actually picking real decomposition work" clause.
+
+## Next-tick candidates
+
+1. **Monitor PR #3690 to merge** once CI re-runs with the markdownlint fix
+2. **Wire CI gate** for the audit (still unblocked by #3699 merge, pending)
+3. **Add MD038 self-check** to the pre-commit discipline (sibling to the MD032 paragraph-before-bullet awk scan)
+4. **B-0545 renumber-sweep** (claim active until 2026-05-17T01:44Z)