diff --git a/docs/hygiene-history/ticks/2026/05/16/0316Z.md b/docs/hygiene-history/ticks/2026/05/16/0316Z.md
new file mode 100644
index 000000000..af0fa86e6
--- /dev/null
+++ b/docs/hygiene-history/ticks/2026/05/16/0316Z.md
@@ -0,0 +1,84 @@
+# Tick 2026-05-16T03:16Z — Otto-CLI
+
+**Surface**: Otto-CLI (Claude Code, Opus 4.7 1M context, autonomous-loop tick)
+**Parent tick**: [2026-05-16T03:10Z](0310Z.md) — baseline mechanism resolves immutability question
+
+## What landed this tick
+
+PR #3698 (0304Z shard) MERGED at 03:09:02Z. PR #3699 (baseline mechanism) attracted 3 new Copilot threads — all real, all fixed via single fixup commit. Pushed `be3998f` to the PR #3699 branch + resolved all 3 threads.
+
+| PR | Action | Result |
+|----|--------|--------|
+| ~~[#3698](https://github.com/Lucent-Financial-Group/Zeta/pull/3698)~~ | (carry-over) | MERGED 03:09:02Z |
+| [#3699](https://github.com/Lucent-Financial-Group/Zeta/pull/3699) | 3 threads → fixup `be3998f` → resolved | OPEN, armed, 0 unresolved |
+| [#3690](https://github.com/Lucent-Financial-Group/Zeta/pull/3690) | armed, awaiting CI | OPEN |
+| [#3701](https://github.com/Lucent-Financial-Group/Zeta/pull/3701) | armed, awaiting CI | OPEN |
+
+## The 3 PR #3699 findings — all real, all fixed
+
+### P1/P2 (line 144): blind cast in `loadBaseline`
+
+`loadBaseline` validated only that JSON parsed to an array; then `data as readonly BaselineEntry[]` cast blindly. Malformed entries (`null`, `line` as string, missing fields) either crashed later in `isInBaseline` (`null.file`) or silently failed to match — converting grandfathered findings to NEW under `--enforce`. Documented behavior says malformed = exit 64.
+
+Fix: `isBaselineEntry` type guard validating each element (`file: string`, `line: integer >= 1`, `target: string`); collect bad entries with index + reason; exit 64 with structured message per entry.
+
+### P1 (line 368): JSON output API mismatch
+
+`baselineMatched: baselineMatched.length` emitted a **number** while docstring described partition into `baselineMatched` vs `newFindings` as **parallel arrays**. Consumers reading the JSON would type-mismatch.
+
+Fix: emit `baselineMatched` as the full array (parallel to `newFindings`); consumers compute count via `.length`. Field is now structurally consistent with the documented partition shape.
+
+## Local verify
+
+| Test | Result |
+|------|--------|
+| Valid baseline | exit 1 (10 grandfathered + 1 transient new) — unchanged |
+| `[null]` baseline | exit 64, "baseline entry [0] invalid: ..." |
+| `line` as string | exit 64 |
+| Missing target | exit 64 |
+| `--json` output | `baselineMatched` is array of 10 (was number); `newFindings` array len 1 |
+| `tsc --noEmit` | exit 0 |
+
+## Audit-script PR lifecycle status
+
+| Step | Tick | PR | Status |
+|------|------|----|--------|
+| Discovery | 3 | n/a | done |
+| Narrow fix per-shard | 3 | #3676 / #3680 | merged |
+| Scanner authored | 7 | #3692 | merged |
+| Filter improvement | 8 | #3692 (fixup) | merged |
+| Quality findings round 1 | 9 | #3692 (fixup) | merged |
+| Quality findings round 2 | 10 | #3692 (fixup) | merged |
+| Baseline mechanism | 11 | #3699 | armed |
+| **Baseline schema + JSON API (round 3 quality)** | **12 (this)** | **#3699 (fixup)** | **armed** |
+| CI enforce wire | next | (gate.yml) | unblocked, pending |
+
+## Operational notes
+
+### Copilot's third pass on the audit
+
+Same PR (across two ticks), three review passes, total findings:
+
+| Pass | Tick | Findings | Severity |
+|------|------|----------|----------|
+| Round 1 | 9 | 4 | 2 P1, 2 P2 |
+| Round 2 | 10 | 2 | 1 P1, 1 P2 (cross-platform) |
+| Round 3 | 12 (this) | 3 | 2 P1, 1 P2 (schema validation + JSON API) |
+
+Cumulative: 9 real findings on the audit script alone, all caught pre-merge. Significant net-positive for the artifact's robustness. The audit is now defensively-coded against bad baselines, has portable cross-platform path handling, validates arg shapes at the boundary, and emits a structurally-consistent JSON API.
+
+### The "third quality round" insight
+
+Each new commit on an open PR triggers fresh Copilot review. Findings from round N might not surface until the round-N-1 commit is in place — the reviewer sees the prior fix, then notices the next-layer concern. This is **iterative refinement at PR scope**, parallel to the iterative refinement at session scope (discovery → narrow-fix → scanner → ...).
+
+The pattern argues for: don't try to predict and pre-empt all findings; ship a reasonable first commit, let review pass over it, fix what's caught, iterate. The "third quality round" is the limit point where new findings approach zero.
+
+### Holding-discipline state
+
+Concrete substrate landed: type-guard validation + JSON API correction. Counter resets per "actually picking real decomposition work" clause.
+
+## Next-tick candidates
+
+1. **Wire CI gate** for the audit (unblocked by #3699 once it merges): add `lint (tick-shard relative-paths)` job to `.github/workflows/gate.yml` invoking `bun tools/hygiene/audit-tick-shard-relative-paths.ts --enforce --baseline tools/hygiene/audit-tick-shard-relative-paths.baseline.json`
+2. **Monitor #3690, #3699, #3701** to merge
+3. **B-0545 renumber-sweep** (claim active until 2026-05-17T01:44Z)