[Hackathon] fix(rerun): a failed deferred-retry chunk discards runs the earlier retry chunks already started and tells the user to rerun them #491

Verified against (commit SHA or released version)

commit 988c035 (v0.14.0)

Discord Username / User ID

uisupersaiyan3

What does this improvement do?

In test rerun --wait (explicit ids or --all), when more than 50 tests stay rate-deferred, the deferred-retry loop dispatches retry chunks sequentially into retryChunkResponses (test.ts:13003-13057). If a later chunk throws on the non-tunnel path, the catch logs and breaks (:13058-13068). The merge of the chunks that already succeeded (accepted, conflicts, notFound, and deferred = newlyDeferred at :13122) sits after the try block, so it is skipped. The tunnel branch of the same catch already calls emitRetryPartial(err, retryChunkResponses); only the ordinary path drops them.

Effect: runs the server accepted in earlier retry chunks are never polled or printed, their ids stay in deferred, the command exits 7, and nextAction says testsprite test rerun <ids> for tests that are already running. Following that hint uses a fresh idempotency key, so it double-dispatches and double-bills.

Trigger: a project over the 50 in-flight cap (so the retry splits into chunks); chunk 0 is accepted, chunk 1 fails with a whole-request 404 (one id lost its replayable run, a documented response shape) or a 5xx after retries.

Details / implementation notes

Repro (vitest, real runTestRerun, mocked fetch, 53 ids, --wait, --output json): initial dispatch accepts 1 and defers 52; retry chunk 0 accepts 50; retry chunk 1 returns 404.

POST batch/rerun #3 size=50 key=...:deferred-retry1:chunk0   (server accepts 50)
POST batch/rerun #4 size=2  key=...:deferred-retry1:chunk1   (404)
EXIT 7  52 test(s) were rate-deferred.
nextAction testsprite test rerun test_001 test_002 ...
server created runs: 51
stdout accepted runs: 1  deferred: 52
runIds visible anywhere in stdout+stderr: 1

Why tests miss it: test.rerun.spec.ts covers retry success, all-deferred and tunnel partials, but no non-tunnel case where a later retry chunk throws after an earlier one succeeded.

Proposed fix (local to the catch, no new dependency): before break, fold the successful retryChunkResponses into accepted / conflicts / notFound / advisories using the same merge code as the success path (extract it into a small local function), and keep in deferred only the failed chunk's ids plus chunks never attempted. Surface the failed chunk's error on stderr as today.

Tests: the 53-id scenario polls and prints 51 run ids, deferred holds only the 2 failed-chunk ids, and nextAction lists only those; existing tunnel-partial tests unchanged.

Not a duplicate: #405 / PR #410 fix the same symptom in the initial dispatch loop (test.ts ~12783-12828); PR #410's diff does not touch the deferred-retry loop. #477 is skipped-array merging in test run --all retries (test.ts ~10264), #471 is exit-code precedence, #426 is a signal during retry sleep. PR to follow once assigned.

Confirmations

• I have searched existing issues and this is not a duplicate.
• I have provided my Discord identity above for reward coordination.