fix(server): retry provider session opens before failing the run

[t3dotgg]

Since #12790, a run fails at the first provider session open error. The effect worker then has nothing to retry, so a short open failure ends the run. SelectionRestart.integration.test.ts fails on t3code/codex-turn-mapping because of this, and every PR stacked on this branch is red.

Fix

• The effect worker passes willRetry to the executor. It is true when a failure will be retried (attemptCount < maxAttempts).
• ProviderTurnStartService.start returns the session open error while willRetry is true. On the last attempt, it sets the run to failed as fix(server): fail V2 runs when provider sessions cannot open #12790 does.
• If there is no caller flag, the fix(server): fail V2 runs when provider sessions cannot open #12790 behavior stays: the run is set to failed.

The #12790 goal stays: a run never stays in starting after the worker stops retrying.

Tests

• New: ProviderTurnStartService.test.ts, "leaves the run starting when a session-open failure will be retried".
• SelectionRestart.integration.test.ts passes again.
• 76 tests pass in ProviderTurnStartService, SelectionRestart, EffectWorker, and RunExecutionService. Server typecheck is clean.

This change was made by Claude Opus 5.5 with Claude Code.

🤖 Generated with Claude Code

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ No successful main baseline artifact is available yet. This run establishes the initial measurement.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	—	4.9 KiB	—	6.8 KiB	✅
Codex	Thread snapshot wire	—	3.7 KiB	—	4.9 KiB	✅
Codex	Live turn WebSocket wire	—	1.1 KiB	—	2.0 KiB	✅
Codex	Live turn WebSocket decoded	—	20.4 KiB	—	29.3 KiB	✅
Codex	Live turn messages	—	1	—	8	✅
Claude	Total thread wire	—	4.9 KiB	—	6.8 KiB	✅
Claude	Thread snapshot wire	—	3.7 KiB	—	4.9 KiB	✅
Claude	Live turn WebSocket wire	—	1.2 KiB	—	2.0 KiB	✅
Claude	Live turn WebSocket decoded	—	20.8 KiB	—	29.3 KiB	✅
Claude	Live turn messages	—	2	—	8	✅

Baseline: unavailable · PR result: f2b8f97 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 106.1 KiB
• Claude decoded thread snapshot: 106.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at f2b8f97

Macroscope's review found this PR approvable — This is a narrowly scoped server bug fix that defers provider-session failure settlement only while the existing effect retry budget remains, while preserving final-attempt behavior and adding regression coverage. It introduces no schema, deployment, product-default, or static-analysis changes.

^{You can add or adjust custom eligibility rules. Learn more.}