The crew's now list says the reviewer is paused when its seat is set to 0, not that there is no reviewer #8778

The crew's now list now says "the reviewer is paused" when the reviewer seat is 0 and the review has lapsed. It still says "no reviewer on the crew" for a crew with no reviewer at all. A reviewer that merely didn't answer still gets "the reviewer did not answer".

The issue's wording didn't quite match how the code works. A pull only waits for a person with "no reviewer" when the crew has no reviewer at all. A reviewer who is paused (seat 0) stays on the crew, so the pull waits for them and shows as waiting for a person only once its review lapses. So "paused" replaces the lapsed wording, not the "no reviewer" wording.

• One source: reviewerPaused(repo) in src/crew/seat_slots.ts is the only place that checks reviewerSlots === 0.
• MCP and API: src/actions/crew_now.ts picks the wording through a new reviewSays. This list feeds the MCP now tool and GET /api/v1/crew/now.
• Page: reviewWhy in src/web/studio/asks.ts takes a paused flag, passed from the ask line and the decision card. Those are the studio's "waits for your review" lines, which I took to be "on the page".
• Test: test/seat-slots.test.ts covers paused, lapsed and no-reviewer in both reviewSays and reviewWhy.
• Not tested end to end: the test calls the wording functions directly. It does not build a lapsed pull in a running app, because that setup is heavy.
• Checks: tsc, scripts/gate.ts, design-lint and the seat-slots, review-waits and mcp-size tests pass. I didn't run the full suite.

I left one thing alone. A pull the reviewer sent back three times ("rounds") still reads "no reviewer on the crew" in the crew's now list. That is wrong, but it isn't this issue.

Closes #8733

The crew's now list now says "the reviewer is paused" when the reviewer seat is 0 and the review has lapsed. It still says "no reviewer on the crew" for a crew with no reviewer at all. A reviewer that merely didn't answer still gets "the reviewer did not answer".

The issue's wording didn't quite match how the code works. A pull only waits for a person with "no reviewer" when the crew has no reviewer at all. A reviewer who is paused (seat 0) stays on the crew, so the pull waits for them and shows as waiting for a person only once its review lapses. So "paused" replaces the lapsed wording, not the "no reviewer" wording.

• One source: reviewerPaused(repo) in src/crew/seat_slots.ts is the only place that checks reviewerSlots === 0.
• MCP and API: src/actions/crew_now.ts picks the wording through a new reviewSays. This list feeds the MCP now tool and GET /api/v1/crew/now.
• Page: reviewWhy in src/web/studio/asks.ts takes a paused flag, passed from the ask line and the decision card. Those are the studio's "waits for your review" lines, which I took to be "on the page".
• Test: test/seat-slots.test.ts covers paused, lapsed and no-reviewer in both reviewSays and reviewWhy.
• Not tested end to end: the test calls the wording functions directly. It does not build a lapsed pull in a running app, because that setup is heavy.
• Checks: tsc, scripts/gate.ts, design-lint and the seat-slots, review-waits and mcp-size tests pass. I didn't run the full suite.

I left one thing alone. A pull the reviewer sent back three times ("rounds") still reads "no reviewer on the crew" in the crew's now list. That is wrong, but it isn't this issue.

All three points and the smaller one are addressed on the same branch. tsc, scripts/gate.ts, design-lint and the seat-slots, review-waits, review-gate and mcp-size tests pass here. I didn't run the full suite.

1. Tests. test/seat-slots.test.ts now builds a green agent pull whose one review went unanswered and whose check is 30 days old, so its review has lapsed. It asserts the wording in three places: crewNow, the MCP now tool, and the studio ask line.

• Seat 8: "the reviewer did not answer".
• Seat 0: "the reviewer is paused".
• Reviewer taken off the crew: "no reviewer on the crew".

The MCP output is checked against crewNow's. I didn't revert the call sites to confirm the new test fails without them.

2. The one source. reviewerPaused now lives in a small leaf module, src/crew/reviewer_paused.ts. It couldn't stay in seat_slots.ts, because review.ts importing it would have made an import cycle. owner_waits.ts and review.ts use it, and so do crew_now.ts, asks.ts and decisions.ts. No hand-written reviewerSlots === 0 is left in src/.

3. Rounds. The "three change requests" case now reads "the reviewer asked for changes three times" on the now list too, so it matches the asks list. It has a unit test on the shared wording, not an end-to-end one, because building a three-rounds pull is heavy.

Shared wording. src/crew/review_says.ts holds reviewSays(wait, paused). The now list calls it directly, and reviewWhy in asks.ts is just that plus ", since <date>". The page, the MCP and the asks can't drift apart.

Late-check wording. A lapsed review that ran late now shows the "review ran late: the check took N min…" text on the now list as well. It showed only on the asks list before.

The check failed because test/seat-slots.test.ts imported openPull twice. Main had added its own import, and my import of the same function sat beside it after the merge. I deleted my copy, which was line 16. Nothing else changed.

After the fix, tsc, scripts/gate.ts and bun test test/seat-slots.test.ts test/mcp-size.test.ts pass here. I didn't run the sharded fast tier locally, so I haven't confirmed it reaches green. The fix is pushed.

None of the three failing tests touch anything this pull changes. I merged the latest main into the branch and pushed; I haven't seen the forge's check re-run, so whether it's green is open.

• "a row wears a dot… under its byte budget" (test/issues-quiet.test.ts) is the only failure I reproduced. The issues page came to 51,705 bytes against a 50,000 budget. Main has since raised that budget to 55,000, so it passes once main is merged in. I confirmed that locally.
• "two previews deploying together…" (took 120 s) and "previews of idle pulls sleep…" are about preview deploys and process sleeping. Nothing in this pull reaches that code, and I didn't run either test, so I can't say whether they fail on main too. If the check fails on them again, treat them as flaky and rerun, or look at them separately.

After the merge, tsc, scripts/gate.ts, test/seat-slots.test.ts and test/issues-quiet.test.ts pass here.