test(server): drop Claude background unit tests the live fixtures cover

[juliusmarminge]

Stacked on #13517. Now that the live fixtures cover Claude background work, the hand-built-frame unit tests that restated the same flows can go. The tests that stay should use frames the CLI actually sends.

Deleted: 8 tests, about 1,030 lines of ClaudeAdapterV2.test.ts

For each test, the adapter rule it guards was reverted and the named fixture failed:

Test	Covered by
projects an authoritative background_tasks_changed roster	claude_background_task_after_root
clears the native-thread roster when a turn is interrupted	claude_background_task_interrupt
drains buffered wake messages into a continuation turn	claude_background_task_wake
terminalizes a continuation turn from a task-notification origin wake result	claude_background_task_wake
wakes and hydrates a subagent that completes after the root turn settled	claude_background_subagent_after_root
re-opens a resumed subagent and hydrates its second result	claude_background_subagent_lifecycle
orders nonempty level, empty level, notification, and continuation drain	claude_background_task_wake
admits only local_bash from a mixed background_tasks_changed snapshot	claude_background_subagent_after_root, claude_background_subagent_lifecycle

The fixtures pick up the assertions those tests made: the continuation detail, and the subagent's run and node attribution. Two sub-checks go without a fixture, because the CLI has not been seen producing them:

• a task_progress after a subagent's notification (none among 5,565 logged progress frames)
• a duplicate local_bash notification (the logged duplicates are all subagents)

Kept, with realistic frames

The remaining background tests cover three kinds of behavior:

• orderings the CLI emits nondeterministically (held frames before task_started, a resume task_started racing past settle, interrupt races)
• failures that cannot be produced on demand (process replacement, failed reopen)
• the idle-release probe

Their frames now match what the CLI sends:

• terminal_reason on results, except a zero-turn task-notification result, which omits it as seen live
• tool_use_id and is_backgrounded on task_started
• tool_use_id on task_notification
• message.id on assistant frames, one content block per frame
• SendMessage input {to, summary, message}, the resume ACK text as recorded, and the resumed notification under the SendMessage tool_use_id
• background Bash ids and texts taken from the claude_background_task_wake recording

The held-frame test still sends a task_started without tool_use_id. The SDK types that field as optional, but no recorded or logged frame has omitted it. The test now says so.

Verification

• vp test run on ClaudeAdapterV2.test.ts, ClaudeAdapterV2.testkit.test.ts, OrchestratorReplayFixtures.contract.test.ts, and scripts/claudeReplayRecordingConfig.test.ts: 140 passed.
• vp test run src/orchestration-v2/testkit/OrchestratorReplayFixtures.integration.test.ts -t claude: 24 passed.
• vp exec tsc --noEmit -p . in apps/server: 0 error TS, 0 warning TS.
• vp run knip:check: passes.
• vp lint on the touched files: clean.
• Not run: repo-wide checks.

Model: Claude Opus 5.5 (Claude Code)

🤖 Generated with Claude Code

---

[macroscopeapp]

Approvability

Verdict: Approved at 419592a

Macroscope's review found this PR approvable — This PR is confined to test-only unit and replay-fixture assertions: it removes duplicated synthetic-frame coverage and strengthens live-fixture checks without touching production code, defaults, or deployment configuration. Its runtime blast radius is limited to test execution.

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: 419592a · Source CI: failure

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.

Dismissing prior approval to re-evaluate 419592a