perf(server): finish runtime messages without full thread reads

[t3dotgg]

Finishing a message or plan could load every message and plan in its thread.

Use keyed message and plan reads, SQL existence checks, and the latest task activity with a usable title. Preserve late completion, buffered replies, implemented-plan metadata, and the existing 500-row compaction history limit. Buffered plan text stays available until its update commits.

Continues #8995 by @rcawston. The original commit was cherry-picked with Ross Cawston's Git authorship preserved. The original PR is unchanged. This continuation fixes its title-loss finding and keeps the newer runtime-context and question-handling behavior on main.

Stacks on #10115, after #10290 and #10123. The base PRs own summary and request lifecycle changes. This PR narrows runtime completion reads.

Verification: 113 focused tests across five server files, server-only typecheck, and changed-file lint and formatting pass on the stack. The saved-title regression fails with the imported query and passes with this fix. No browser or provider process was run.

The existing terminal-turn pending-question lifecycle read is unchanged. No wire schema, database migration, or UI change.

Created with GPT-6 Astra in Codex. Follow-up review and restack with Codex.

Note

Replace full thread-detail reads with targeted projection queries in ProviderRuntimeIngestion

• Replaces thread-detail hydration in provider runtime event processing with exact repository lookups for messages, proposed plans, and task activities.
• Adds ProjectionThreadProposedPlanRepository.getByPlanId, ProjectionThreadMessageRepository.hasAssistantMessageForTurn, and ProjectionThreadActivityRepository.getLatestTaskActivity; extends list activities with kind filters and limits.
• Fixes blank buffered proposed-plan finalization to clear the buffer without creating a row, allowing a later nonblank completion to replace the plan while preserving creation and implementation metadata.
• Compaction token history now queries task activities with kind filters and a 500-row bound instead of loading full thread detail.
• Behavioral Change: task-title recovery now selects the latest persisted task activity with a nonblank title rather than relying on in-memory thread search; listProjectionThreadActivityRows applies newest-first limiting before returning ascending order.

^{Macroscope summarized d8f7c40.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.6 KiB	13.5 KiB	−40 B (−0.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.0 KiB	7.0 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.5 KiB	−41 B (−0.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	57.1 KiB	57.0 KiB	−88 B (−0.2%)	66.4 KiB	✅
Codex	Live turn messages	10	8	−2 (−20.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.6 KiB	+76 B (+0.5%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.0 KiB	7.0 KiB	+4 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.6 KiB	+72 B (+1.1%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.8 KiB	57.9 KiB	+88 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	8	10	+2 (+25.0%)	21	✅

Baseline: eb8ed80 · PR result: d8f7c40 · 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: 113.8 KiB
• Claude decoded thread snapshot: 114.5 KiB

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

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR is a substantial production-path refactor that replaces full thread hydration with multiple targeted persistence queries and changes runtime message, plan, task, and compaction handling. The unresolved high-severity request-response classification finding also requires human attention.

^{Adjust the Minimum Blocking Severity for this repo — including turning it Off — in Settings. You can add or adjust custom eligibility rules. Learn more.}

Dismissing prior approval to re-evaluate 857ef05

[cursor]

Bugbot is paused — on-demand spend limit reached

Bugbot uses usage-based billing for this team and has hit its on-demand spend limit.

A team admin can raise the spend limit in the Cursor dashboard, or wait for the next billing cycle to continue.