Comment-only. `tsc` 0 errors, `census --strict` and
`check-fnxc-future-dates` exit 0.
## Claimed with the new tool
First use of `scripts/check-file-claimed.mjs` (#3175) to pick work
instead of guessing:
```
CLAIMED packages/engine/src/scheduler.ts #3177, #3142
CLAIMED packages/core/src/task-store/audit-ops.ts #3165
UNCLAIMED packages/core/src/mission-store.ts
UNCLAIMED packages/core/src/task-store/task-id-integrity.ts
```
Two of the four files I would have reached for were already taken — by
PRs whose branch names give no hint they touch those paths. That is the
collision this phase paid for five times, answered in one command. I
took `mission-store.ts`; `task-id-integrity.ts` is still free.
## Census 13 → 12
Reclassification, not conversion — the line is unchanged.
## Verified the blocker rather than deferring to it
The site carries an audited note: the sync `MissionStore` reaches
`this.db.prepare`, and `getMissionStoreImpl` returns the
`AsyncDataLayer`-backed `AsyncMissionStore` under PostgreSQL, so the
class is unreachable in the shipped backend.
I checked that independently instead of accepting it —
`async-mission-store.ts:168` states the same routing from the other
side. **That check exists because of #3129**, where a note I had
accepted as settled ("blocked on a per-neighbour flag map that does not
exist") turned out to name the wrong variable, and the file was
convertible all along. I had publicly argued it should stay counted.
So the rule I am applying: a documented blocker gets marked only after
its named obstacle is confirmed from a second source. Here it held; on
`taskRevert.ts` it did not.
## Related, and still open
`merge-queue-ops-2.ts` carries a note of the same shape that does
**not** survive this check — it names two ways to convert (make the path
async, push the trait read down) and misses the one that worked twice
this phase: thread the resolved lanes in from a caller that already
awaited them, as #3112 and #3118 did for `executor.ts`. Its sibling
`taskStillInReview(projectId, reviewColumns)` already takes lanes from
its caller. Worth a real look rather than a marker.
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2cb5cab595