perf(mobile): reuse the settled sort when settled rows are unchanged

[juliusmarminge]

On mobile Home, five streaming threads cause about 660 shell thread.updated events in 15 s. Each one rebuilds the thread list. Every rebuild re-sorted all settled shells (1204 in the fixture) with sortSettledThreads and re-parsed their timestamps, even though the streams only change unsettled rows. In the Hermes profile this sort accounted for 186 and 224 inclusive samples per capture.

buildThreadListV2Items now keeps one entry: the last settled input array and its sorted output. If the next settled array has the same length and the same elements (===) in the same order, it reuses the output. Otherwise it calls the unchanged sortSettledThreads and replaces the entry. This is safe because shells are immutable (upsertById replaces a row, and the scopedThread WeakMap keeps row identity) and the comparator reads only element fields. now, search, queued keys, and settlement gating all run before the settled array is built, so the identity check covers them. Home and the split-view sidebar share the entry, so a different scope is a cache miss and recomputes, never a stale hit. Retention is bounded to two arrays of references the shell store already holds. Paging previously did pagedSettled.push(selected), which could mutate the sort output. It now builds [...limited, selected], so the cached array is never mutated. Web is unchanged.

Measurement

Two serialized A/B pairs (before 1, after 1, before 2, after 2) on an iOS simulator release build (Hermes, no Metro). Native binary is identical, only the JS bundle is swapped. Before: bundle 068d9912… (#16330 head). After: bundle 92c76eea… built from 176294b18d. Its composed source map differs from the before map only in threadListV2.ts, and that file is byte-equal to the committed blob. Each run: Home visible, five synthetic 20 Hz streams for 15 s on unsettled threads, a 1210-shell fixture (Home shows "1204 settled threads"), and the same restored fixture and client DB. All 5 streams completed 300/300 in every run.

Capture	App CPU (s)	sortSettledThreads incl.	buildThreadListV2Items incl.	self dateParse	self arrayPrototypeSort	GC	Active / total samples	Active %, 0–16 s
before 1	4.55	186	248	46	75	253	2259 / 14539	18.45
after 1	4.34	0	52	0	2	229	1904 / 12549	17.30
before 2	4.45	224	274	57	86	275	2424 / 15052	18.70
after 2	4.28	0	71	3	0	274	2290 / 15007	17.72

App process CPU fell by 4.6% and 3.8% in the two pairs (0.01 s ps resolution). The settled sort is gone from both after captures. Sample counts are inclusive and overlap, so they must not be summed. After 1 has fewer total samples because it had 57 sampler gaps over 20 ms (about 2.0 s in total), all after 16.4 s, when the streams had ended. That makes its trace span 23.83 s, against a profiler window of about 22.59 s. Its raw active count is therefore not directly comparable. The last column normalizes to the gap-free 0–16 s window, where every run's largest gap is ≤14.4 ms.

This is a narrow workload: many settled threads with streams on unsettled ones. Nothing here measures latency, frames, or battery, and nothing ran on a physical device. Aggregate evidence (no traces or client data): settled-sort-evidence.json.

Tests

threadListV2.test.ts adds 5 tests under "buildThreadListV2Items settled sort reuse", for 102/102 passing. They cover the baseline order; a re-sort after a settled row is replaced, added, or removed; a re-sort when tied rows are reordered; a skipped sort when only an unsettled row changes (a Date.parse spy); and alternating environment scopes. Disabling the identity check fails 3 of them. Disabling reuse fails 1.

Opus 5.5 via Claude Code in T3 Code.

🤖 Generated with Claude Code

---

[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	—	5.0 KiB	—	6.8 KiB	✅
Codex	Thread snapshot wire	—	3.8 KiB	—	4.9 KiB	✅
Codex	Live turn WebSocket wire	—	1.2 KiB	—	2.0 KiB	✅
Codex	Live turn WebSocket decoded	—	20.9 KiB	—	29.3 KiB	✅
Codex	Live turn messages	—	2	—	8	✅
Claude	Total thread wire	—	5.0 KiB	—	6.8 KiB	✅
Claude	Thread snapshot wire	—	3.8 KiB	—	4.9 KiB	✅
Claude	Live turn WebSocket wire	—	1.2 KiB	—	2.0 KiB	✅
Claude	Live turn WebSocket decoded	—	21.2 KiB	—	29.3 KiB	✅
Claude	Live turn messages	—	2	—	8	✅

Baseline: unavailable · PR result: 35ab4d9 · 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.

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR adds production-path memoization that automatically skips sorting and timestamp parsing for unchanged settled-thread rows, with behavior guarded by identity checks and focused tests. Because it changes when significant list-processing work executes on the mobile request path, an additional human review is warranted despite the localized diff.

No code changes detected at 35ab4d9. Prior analysis still applies.

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

[coderabbitai]

Important

Review skipped

We couldn't safely recover the incremental review. No full review was started, and the last reviewed checkpoint was preserved. Retry later, or explicitly request a full review by commenting @coderabbitai full review.

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

• Configuration used: Repository: pingdotgg/t3code/.coderabbit.yaml
• Review profile: CHILL
• Plan: Team
• Run ID: e689f482-359a-4c5c-be0f-0c174ec080ba

📥 Commits

Reviewing files that changed from the base of the PR and between 7bff9d5 and 176294b.

📒 Files selected for processing (2)

• apps/mobile/src/features/threads/threadListV2.test.ts
• apps/mobile/src/features/threads/threadListV2.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 4 remain after this review.

---

📝 Walkthrough

Walkthrough

The thread list now reuses settled-row sorting when its input references are unchanged. It also keeps a selected settled thread visible when that thread falls beyond the page limit.

Changes

Settled Thread List

Layer / File(s)	Summary
Settled sorting and cache validation apps/mobile/src/features/threads/threadListV2.ts, apps/mobile/src/features/threads/threadListV2.test.ts	A one-entry cache reuses the sorted settled rows when the input references match. Tests cover newest-first ordering, input changes, stable ordering for equal timestamps, parse counts, and alternating environment scopes.
Selected thread paging apps/mobile/src/features/threads/threadListV2.ts	A selected thread beyond the limit-sized settled prefix remains visible. The prefix continues to determine the displayed settled page.

Priority: ⬇️ Low

Estimated code review effort: 3 (Moderate) | ~20 minutes

Change: Refactor

Suggested reviewers: t3dotgg

Merge Risk: ⚪ Minimal · up to 17629

This change skips re-sorting settled threads when their rows are unchanged, which reduces CPU use on Home while threads stream. No concrete defect was found, so the merge risk is low.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 28.57% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 7 functions across 2 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely identifies the main change: reusing the settled-thread sort when settled rows are unchanged.
Description check	✅ Passed	The description explains the problem, implementation, and focused verification in detail. It does not include the required Scope and approval section or explain why prior approval is not needed.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches 💡 2

📝 Generate docstrings 💡

• Commit to this branch
• Create a new PR

🛠️ Fix failing CI checks 💡

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

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

---

_{Comment @coderabbitai help to get the list of available commands.}