fix(server): a retried pull-request-state delivery never fails with 500

[tusharbhardwaj-bk]

Problem

A stress run on expbkt3 (10 concurrent pull-request-state writers against 5 threads on one branch) got HTTP 500 on 27 of 160 writes:

pull request state write failed
  _tag: 'OrchestrationCommandPreviouslyRejectedError'
  commandId: 'bridge:pr-state:stress-1-0:perf-stress-thread-…:branch'
  detail: 'Orchestration command invariant failed (thread.pull-request.sync): thread … changed before pull request discovery'

The endpoint (#217) treats a thread that changed between its read and the dispatch as stale. But a retry of the same delivery reuses the same command id (bridge:pr-state:<deliveryId>:<threadId>:…), and the engine answers a command it already rejected with OrchestrationCommandPreviouslyRejectedError. That surfaced as a 500. The bridge retries failed deliveries, so it would have retried that delivery forever. The thread.pull-request-link.sync dispatch had no race handling at all.

Fix

Both dispatches now report either OrchestrationCommandInvariantError or OrchestrationCommandPreviouslyRejectedError as ignored: stale with HTTP 200. That matches the contract: "a write that lost to the thread's own changes is ignored".

Evidence

• New test treats a retried delivery whose update was rejected as stale, not a failure: it pre-records a rejected command under the retry's command id, then applies the delivery. It fails without the fix and passes with it. The suite passes (7).
• vp run typecheck in apps/server and lint are clean.

After it deploys I'll re-run the stress test on expbkt3 and expect 0 errors.

Model/harness: Claude Opus 5.5 via Claude Code in T3 Code.

🤖 Generated with Claude Code

[github-actions]

Thread transfer impact

⚠️ The latest CI run did not produce a thread transfer result for ee20f2a.

This comment will update automatically after the next completed run.