test(server): make the provider-switch resume-failure test reach the fallback

[juliusmarminge]

"uses portable fallback when native resume fails after a provider switch" never reached the fallback. The shared Codex session kept its native thread loaded while the Claude turn ran, so returning to Codex reused it warm and never called resumeThread. Flipping failResume: true to false still passed.

What changed

• Before returning to Codex, the test stops the Codex provider session with provider-session.detach, the same command the client's stop-session action sends, and drains the effect worker. The return now has to resume the native thread, the fake fails that resume, and ProviderTurnStartService takes the portable fallback.
• The assertions now check what the fallback actually produces, instead of warm reuse:
  • a fresh native Codex thread (codex:<thread>:1) bound to the same provider-thread row, with the fake's native-thread generation counter at 2;
  • a third handoff (full_thread_summary, runs 1-2) delivered inline to that fresh native thread, next to the Claude delta;
  • a third provider_handoff transfer with status resolved_portable and strategy portable_context;
  • a return prompt that re-sends the earlier Codex turn ("codex before switch"). The old assertion that it was absent was the warm-reuse expectation, so it is replaced.

With the path actually exercised, the fallback behaves correctly. No production code changed.

Verification

• vp test run src/orchestration-v2/testkit/ProviderSwitch.integration.test.ts: 65/65 pass.
• Mutation checks on the rewritten test (each fails, then passes again after reverting):
  • failResume: false: fails with expected 1 to equal 2 (native thread generation). With that assertion removed it still fails on expected 'codex:thread:provider-switch:0' to equal 'codex:thread:provider-switch:1'.
  • Fallback removed in ProviderTurnStartService (propagate the resume failure instead of replacing the native thread): fails. The run retries the resume and the test times out waiting for idle.
  • Fallback handoff dropped (effectiveHandoffs = [handoff, ...] skipped): fails on the handoff list assertion.
  • Detach step removed (the old warm-reuse setup): fails with expected 1 to equal 2.
• vp lint and vp fmt on the changed file: clean. vp exec tsc --noEmit -p . in apps/server: no error TS.
• Not run: repo-wide checks.

Model: Claude Opus 5.5 (Claude Code)

🤖 Generated with Claude Code

---

[macroscopeapp]

Approvability

Verdict: Approved at 7cbad6d

Macroscope's review found this PR approvable — This PR is limited to an ignored integration-test file and refines the test setup and assertions for an existing provider-switch fallback scenario. It introduces no production runtime, product-default, schema, infrastructure, security, billing, or static-analysis changes.

Notes:

• No code objects were reviewed. Approvability was decided on eligibility alone.

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

[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: 7cbad6d · 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.