fix(server): shell reads no longer scan every runless item per failed thread

[juliusmarminge]

The V2 preview froze on startup for minutes at 100% CPU: the server never answered, so the app showed no providers. #13877 added a per-thread lookup of the failed run's error item to the thread shell query. When that run has no root node, SQLite matched items by `node_id IS NULL` through the node index. Imported history leaves tens of thousands of items with a null `node_id`, so the query walked all of them once for every failed thread. A real database (1,918 threads, 304k turn items) spent 91 of 92 profiled seconds inside this synchronous query.

Both failure lookups now use the thread and run index.

• Timing: on a copy of real data with 1,900 failed threads and 67k runless items, the shell query went from 185 s to 76 ms, with identical rows.
• Test: the new plan test in `ProjectionSettlement.test.ts` captures the real shell query and checks that both item lookups use `turn_items_thread_run_idx`. It fails without this change.
• Checks: ProjectionStore, settlement, control-read and runtime-layer tests pass (89), and the server typecheck is clean.

🤖 Generated with Claude Code

---

[macroscopeapp]

Approvability

Verdict: Approved at d7aac6c

Macroscope's review found this PR approvable — This is a narrowly scoped server bug fix that keeps shell failure results unchanged while constraining both lookups to the current thread and an existing index, preventing expensive scans of unrelated runless history. A targeted regression test verifies both query plans, and no defaults, schemas, capabilities, or sensitive areas are changed.

^{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: d7aac6c · 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.