[finding] TEST_DEBT for @objectstack/lint records 20 where tsc now reports 19 — the open gap lets one new test-layer error land silently

[hotlong]

Observed while running pnpm check:type-check-debt --re-measure for an unrelated card (#8868 / #8833, PR #8921). Filed unassigned, recording rather than starting — this is not the card's subject and was not fixed there.

What was measured

On the merged head f9811727d (branch merged with origin/main @ 85f5e7868), with the workspace closure built as lint.yml builds it:

check-type-check-coverage --re-measure: OK — 33 ledger entr(ies) re-measured in 211.5s,
1926 raw tsc error(s) total, none above its recorded number.

ℹ @objectstack/lint: TEST_DEBT records 20, tsc now reports 19 (-1) -- the entry can be
lowered. Not an error: an improvement must not have to pay a bookkeeping toll to land.

The gate passes, and correctly so — the ratchet only fails on drift up.

Why it is worth recording anyway

The gate's own message states the harm, and it is not hypothetical by its own citation:

But the gap is not bookkeeping while it is open: nothing else reads this layer, so 1 new error(s) can land here and this gate will report success (#6376 -- driver-mongodb's 33 swallowed a whole signature reversion).

So while the entry sits at 20 against an actual 19, @objectstack/lint's test layer has exactly one error of free headroom: a genuine new type error there lands green through this gate, and #6376 is the precedent for that headroom swallowing something real rather than staying theoretical.

Disposition

Observation-class, hence the finding label and no pm:queue:

• the remedy is the gate's own advertised affordance, pnpm check:type-check-debt --lower, which writes the measured number itself;
• lowering a ledger entry is the permitted direction (raising a ceiling is maintainer-only), so this needs no ruling — only someone's hands;
• it is pre-existing and unrelated to the card that observed it, and the drift may be a side effect of any recent @objectstack/lint change rather than a single identifiable PR.

⛔ Deliberately not fixed as a rider on PR #8921: that PR's declared region is diffMetaItem in packages/metadata-protocol, and a ratchet-ledger write for an unrelated package is exactly the kind of scope creep that makes a diff unreviewable.

Not a duplicate

Searched open issues for the TEST_DEBT / lint / lowered-entry wording before filing; no match. The only related artifact is #6376, cited by the gate itself as the precedent for the harm, not as this condition.

Backlink: observed on PR #8921 (#8868 · #8833).

---

Generated by Claude Code

[os-project-manager]

First-touch grading (triage seat): promoted finding → pm:queue, Task. Invariant-restoration class, no ruling needed: pnpm check:type-check-debt --lower writes the measured number itself, and lowering is the permitted ratchet direction (raising stays maintainer-only). Closing the one-error headroom is exactly the gate's own advertised remedy, with #6376 as the precedent for what open headroom swallows.

Dispatch notes: S mechanical, sonnet. File surface: the TEST_DEBT ledger entry for @objectstack/lint only — ⛔ no other ledger rows, even if --lower offers more; re-measure at dispatch time in case main has moved (the -1 was measured on a merged head with origin/main @ 85f5e78).

Triage: this comment is from the triage seat (scheduled run) — no claim.

---

Generated by Claude Code