perf(mobile): speed up large Android threads

[shivamhwp]

Large Android threads could hydrate unbounded database history, retain multi-megabyte tool payloads, repeatedly parse timestamps while rebuilding the feed, and wait on a failing native highlighter probe. Cold thread routes also replaced an empty loading toolbar with the hydrated title/actions, causing a visible layout shift.

This change:

• bounds database-backed detail history to the live projector limits while preserving unbounded checkpoint history for operational turn-diff lookup;
• compacts superseded context-window telemetry before applying the activity limit, retaining unresolved rows;
• projects oversized command/MCP payload previews without changing persisted data;
• parses feed timestamps once and uses recycling-aware Android rows with content-keyed feedback state;
• renders the Android loading/unavailable toolbar at final geometry, including a working cold-route back/Home fallback;
• uses the documented JavaScript highlighter by default instead of waiting for a failing native probe.

Production-shaped Android emulator A/B, using the same copied thread data and no screen recording:

• 478 messages / 6,696 activities: cached snapshot 3.01 MB → 1.05 MB; load RSS growth ~204 MB → ~21 MB; p50/p95 42/150 ms → 26/81 ms; janky frames 36.4% → 30.4%.
• 102 messages / 1,649 activities: cached snapshot 775 KB → 356 KB; p95 81 ms → 57 ms.
• 4-message control: p50/p95 remained 17/57 ms.
• highlighter initialization: ~5.2 s native failure/fallback → ~0.64 s direct JavaScript initialization.
• loading and hydrated Android header title/action bounds were visually verified to remain fixed.

Verification:

• 52 focused server/mobile tests across projection queries, projector behavior, payload projection, highlighter selection, and thread activity
• mobile and server typechecks
• targeted formatting and lint
• Android API 36 preview-build passes across tiny, medium, giant settled, and active copied threads
• independent Claude Code/Fable and Codex performance audits; the checkpoint-history coupling they found was fixed before this update

Implemented with GPT-5.6 Sol in the T3 Code Codex harness, with an independent Claude Code/Fable review.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: c9ccfcfa-b599-4e95-84d2-3e7eebb413ee

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

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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