[juliusmarminge]

The check for the latest agent event without a thread (latestAgentSequence()) can be planned onto the covering per-stream index idx_orchestration_events_agent_stream_sequence. To find the max, that plan scans every agent event. Each streamDomainEvents subscriber runs this query at boot, and so do Orchestrator.layer and ProjectionMaintenance. Fleet telemetry for 2026-11-01 → 11-02 shows 1,211 runs over 150 ms in a day on servers built from main. They took about 520 s in total, 91% of it executing rather than waiting on locks, with a max of 740 ms.

The fix pins idx_orchestration_events_application_sequence (application_event_version, sequence), so MAX(sequence) seeks to the end of the version-2 range. The scoped (per-thread) branch is unchanged.

Scope and approval

Maintainer-authored fix from the 2026-11-02 fleet trace audit. It changes the plan for one read query only and no behavior.

Evidence

Read-only connection to a real 12.4 GB database (1.45M events):

Plan	Cold	Warm
Covering stream index (the slow plan seen in traces)	9,281 ms	182 ms
Pinned application_sequence index (this PR)	0.14 ms	0.05 ms

SEARCH orchestration_events USING INDEX idx_orchestration_events_application_sequence (application_event_version=?)

The existing OrchestrationEventStore.sequence.test.ts covers the results with project, legacy and other-thread events mixed in, and it passes unchanged. No screenshots, because this changes no UI.

🤖 Generated with Claude Code (Claude Opus 5.5)

[macroscopeapp]

Approvability

Verdict: Approved at 75959c0

Macroscope's review found this PR approvable — Pins an existing index to speed up the unscoped agent-event high-water query without changing its filters or results. The read-only change adds no schema changes or data mutations, and reported real-database measurements show a substantial performance improvement.

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

[coderabbitai]

No actionable comments were generated in the recent review.

Recent review info

Run configuration

• Source mode: Auto
• Repository: pingdotgg/t3code
• Branch: perf/latest-agent-sequence-index
• File: .coderabbit.config.ts
• Commit: 75959c0
• Layers: Repository TS config
• Review profile: CHILL
• Plan: Team
• Run ID: 2e9ebbd1-e86e-479f-953d-146694bf3b13

Commits

Reviewing files that changed from the base of the PR and between 60aec28 and 75959c0.

Files selected for processing: 1

• apps/server/src/persistence/OrchestrationEventStore.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 8 remain after this review.

Walkthrough

latestAgentSequence now forces the application-sequence index when no threadId is supplied. When a threadId is supplied, it continues to use the thread-specific index.

Changes: Sequence index selection

Layer / File(s)	Summary
Select the sequence index apps/server/src/persistence/OrchestrationEventStore.ts	When threadId is undefined, latestAgentSequence uses idx_orchestration_events_application_sequence. The thread-specific index branch is unchanged.

Estimated code review effort: 2 (Simple) | ~5 minutes

Change: Refactor

Suggested reviewers: maria-rcks

Merge Risk: Minimal · up to 75959

The optimization preserves agent high-water results and introduces no identified merge-blocking risk.

Pre-merge checks: 4

Check name	Status	Explanation
Title check		The title clearly identifies the main change: pinning the index for the unscoped agent event high-water query. It is concise and uses a conventional commit format.
Linked Issues check		Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check		Check skipped because no linked issues were found for this pull request.
Description check		The description explains the performance problem, the index-pinning change, scope, approval basis, benchmark evidence, and test results. It does not use the template headings exactly: it uses an openi…

Full details: Description check

Explanation

The description explains the performance problem, the index-pinning change, scope, approval basis, benchmark evidence, and test results. It does not use the template headings exactly: it uses an opening summary instead of Problem and Change, and Evidence instead of Verification. The required information is still present and sufficiently specific.

Finishing Touches

Generate docstrings

• Commit to this branch
• Create a new PR

Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

[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	5.0 KiB	5.0 KiB	+6 B (+0.1%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.3 KiB	+6 B (+0.5%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	21.6 KiB	21.6 KiB	0 B (0.0%)	24.4 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.3 KiB	1.3 KiB	0 B (0.0%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.9 KiB	21.9 KiB	0 B (0.0%)	24.4 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: 60aec28 · PR result: 75959c0 · 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: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

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