A01 native comparability — round 4, [#33042](https://git.cafe/pingdotgg/t3code/pull/33042) ([mirror #33042](https://github.com/pingdotgg/t3code/pull/33042)).
**Starting state passes: zero T16 pixels on both independent takes.** The prior one-physical-pixel settled-row offset disappeared with matched launch/Settled expansion/scroll-to-top. Cause classification: capture variance. No reproducible extra settled-row code change. Identical mobile source hashes and the unchanged slim layout at [thread-list-v2-items.tsx:1472](https://github.com/pingdotgg/t3code/blob/28ee80fa3a26b4663950345034f971887d0ad2f8/apps/mobile/src/features/threads/thread-list-v2-items.tsx#L1472) support this; the exact round-3 gesture that produced the fractional offset was not isolated.
Same leased iPhone 17 Pro, iOS 27.2, portrait 402×874 points / 1206×2622 pixels, Large dynamic type / fontScale 1, dark, fixed display-relative clock 2026-11-10 12:00 UTC. Both raw home indicators are hidden. Only the top status-bar rectangle [0,0,1206,180] is replaced identically from before-start-take1-a.png on all public stills/recordings. No row registration or bottom-chrome masking.
| State / device / theme | Before | After | Diff0 (amplified) | DiffT16 |
|---|---|---|---|---|
| Settled start · iPhone 17 Pro · dark | 
[4c361199f4](https://git.cafe/pingdotgg/t3code/commit/4c361199f47f6ea32a2c14835df392640d0f7c63) | 
[28ee80fa3a](https://git.cafe/pingdotgg/t3code/commit/28ee80fa3a26b4663950345034f971887d0ad2f8) | 
[4c361199f4](https://git.cafe/pingdotgg/t3code/commit/4c361199f47f6ea32a2c14835df392640d0f7c63) → [28ee80fa3a](https://git.cafe/pingdotgg/t3code/commit/28ee80fa3a26b4663950345034f971887d0ad2f8) | 
[4c361199f4](https://git.cafe/pingdotgg/t3code/commit/4c361199f47f6ea32a2c14835df392640d0f7c63) → [28ee80fa3a](https://git.cafe/pingdotgg/t3code/commit/28ee80fa3a26b4663950345034f971887d0ad2f8) |
| After Un-settle · iPhone 17 Pro · dark | 
[4c361199f4](https://git.cafe/pingdotgg/t3code/commit/4c361199f47f6ea32a2c14835df392640d0f7c63) | 
[28ee80fa3a](https://git.cafe/pingdotgg/t3code/commit/28ee80fa3a26b4663950345034f971887d0ad2f8) | 
[4c361199f4](https://git.cafe/pingdotgg/t3code/commit/4c361199f47f6ea32a2c14835df392640d0f7c63) → [28ee80fa3a](https://git.cafe/pingdotgg/t3code/commit/28ee80fa3a26b4663950345034f971887d0ad2f8) | 
[4c361199f4](https://git.cafe/pingdotgg/t3code/commit/4c361199f47f6ea32a2c14835df392640d0f7c63) → [28ee80fa3a](https://git.cafe/pingdotgg/t3code/commit/28ee80fa3a26b4663950345034f971887d0ad2f8) |
| Comparison | Diff0 pixels / bbox | T16 pixels / bbox |
|---|---|---|
| Starting take 1 | 173 / [87,2373,1137,2559] | 0 / empty |
| Starting take 2 | 173 / [87,2373,1137,2559] | 0 / empty |
| Endpoint | 63,733 / [53,505,1154,963] | 48,562 / [53,505,1154,960] |
| Each of six repeated A/B still pairs | 0 / empty | 0 / empty |
Bboxes are [left,top,right,bottom), physical pixels. Starting-state row pixels are exactly identical even at Diff0. The 173 starting-state pixels are preserved native bottom-toolbar glass-edge variance, channel delta 1–2; none reaches T16. Endpoint Diff0 and DiffT16 are entirely confined to the primary row's group/position, Waiting glyph → age, emphasis/provider-mark opacity, and the resulting shift of the Settled shelf/remaining row. Zero endpoint changes in unrelated chrome.
**What else changed:** status-bar normalization is identical and disclosed above; starting-state native glass edges differ only at levels 1–2. Both raw home indicators match (hidden). Recording command/animation durations vary. There is no reproducible settled-row offset or additional code-driven visual change.
New round-4 recordings: [Before Un-settle](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c397cef5f300bc87/before-unsettle-public.mp4) (9.43s; menu ~5–7s, endpoint ~9s), [After Un-settle](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/bef95efa5fb4d208/after-unsettle-public.mp4) (16.97s; menu ~9–15s, endpoint ~16s). Before long-press requested 10:59:31.194 UTC, Un-settle 10:59:35.346; after long-press 11:05:05.693, Un-settle 11:05:14.556. See [timestamps](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/bd4e68df5fe182d8/timestamps.json) for complete UTC request/completion times. Native export omits idle frames, so these are action proof, not performance timings.
[Scripts](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7d61cd54649d364e/scripts.md) · [Recipe, steps and limitations](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f1f494405a34414a/workflow.md) · [Counts and boxes](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/94497c6e1c6d6583/verification.json) · [All raw/normalized captures, repeated takes, all diffs, recordings, scripts and source provenance](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/28049f3d8c432615/A01-native-round4-bundle.tar.gz). The archive contains manifest.json with SHA-256 for every included asset.
**Unverified:** exact 500 ms sensor-to-sensor cadence. The driver pauses 500 ms after the first screenshot request completes, with additional remote capture/transport latency. All normalized repeated screenshots are identical. Other devices/scenarios were outside this native A01 round. Evidence-only: no tests rerun.
Cleanup verified: both synthetic connections removed, native catalog empty, lease closed, four owned process groups stopped, no owned port listeners/processes, display-clock source restored, both created worktrees removed. Existing review worktree and PR body untouched. Disk available: 101 GB.
AI-assisted; verified with the tests and tools listed above.
FRICTION:
- tool | iOS launch confirmation was left open despite a successful open response; alert accept hit ALERT_DEADLINE_EXCEEDED. Screenshot-guided coordinate acceptance was required. | Reliable native launch-confirmation handling.
- infra | The leased device became unavailable at 11:05:52 UTC during endpoint capture; recovered on the first retry after a full minute at 11:07:04 UTC. | Stable pool session routing and capacity.
- tool | Screenshot calls use a 500 ms pause between completed requests; remote capture/transfer latency prevents claiming a 500 ms sensor-to-sensor interval. A daemon batch honored a 500 ms wait but did not download its nested screenshot files. | Nested batch screenshot artifact delivery and sensor timestamps.
All 15 uploaded assets were downloaded again with curl and matched their local SHA-256. The inline page preview timed out twice; html_render displayed the static comparison. This affected presentation only; accepted images/diffs were reviewed locally.
Verbatim `ghdrop upload -m` output:
[before-unsettle-public.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c397cef5f300bc87/before-unsettle-public.mp4)
[after-unsettle-public.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/bef95efa5fb4d208/after-unsettle-public.mp4)
[A01-native-round4-bundle.tar.gz](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/28049f3d8c432615/A01-native-round4-bundle.tar.gz)
[scripts.md](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7d61cd54649d364e/scripts.md)
[workflow.md](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f1f494405a34414a/workflow.md)
[verification.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/94497c6e1c6d6583/verification.json)
[timestamps.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/bd4e68df5fe182d8/timestamps.json)