# Evidence r2 — terminal drawer follow
Captured **4/4 requested rows**, R01, R02, R07b, R09b, against real isolated web app servers for [#32826](https://git.cafe/pingdotgg/t3code/pull/32826) ([GitHub #32826](https://github.com/pingdotgg/t3code/pull/32826)). Before [20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718); after [f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a). No PR worktree edits or commits.
## Rendering conditions
Every capture in this report: Playwright 1.60.0; HeadlessChrome 148.0.7778.96; Linux x86_64; no device emulation; viewport 1440×900; DPR 1; font scale 1; root 16 px; dark theme; normal motion; panel animation 200 ms; loaded fonts. Browser Date frozen at 2026-11-10T12:00:00.000Z, while performance clock and real stream remain live. Screenshots preserve the whole 1440×900 surface. Videos are 60 fps H264, constructed from timestamped CDP PNG frames; they repeat frames between browser paints and do not imply 60 distinct captured paints/second.
Fresh databases were created by each checkout's own server with T3CODE_DEV_SEED=0. No live userdata or another lead's state was read or copied. Same synthetic Replay transcript and preparation on both sides; initial active thread content/scroll state match. Only the synthetic Replay showcase transcript was replaced in the two disposable checkouts; app and server implementation are the specified commits.
## Reading the measurements
Distance is max(0, contentLength − viewportHeight − scrollTop), in CSS pixels, from Legend list state. The post-mount ResizeObserver series is the primary resize proof. “Height-change peak” selects deliveries where height differs from the previous delivery. “All RO peak” retains unchanged-height mount deliveries too. rAF can run before Legend resize work and the PR's queued microtask, so a large rAF value is not itself proof of a painted after-state defect. All raw rAF and RO data remain available.
streamChars counts the visible current assistant DOM text, not the stale timeline-entry props: message rows subscribe to a live buffer independently. The sampler records fiberStreamChars separately. Continuation growth plus changing content height and following=true demonstrates the live stream and resumed follow. Working=true was retained throughout every accepted recording.
| Row | Before height-change peak | After height-change peak | All RO peak before / after | RO deliveries before / after |
|---|---:|---:|---:|---:|
| R01 | 593 px | 0 px | 593 / 0 px | 89 / 75 |
| R02 | 589 px | 0 px | 589 / 30 px | 13 / 8 |
| R07b | 179 px | 0 px | 179 / 0 px | 13 / 11 |
| R09b | 643 px | 0 px | 643 / 55 px | 21 / 23 |
R02 after has one 30 px unchanged 848 px delivery as content grows during drawer mount; R09b has one 55 px unchanged 848 px delivery. Those values are retained, rather than reporting every observer delivery as zero. During pure content growth, both before and after retain smooth following, with sampled waves up to 112 px that decay to zero. This is distinct from the resize defect.
[metrics.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/3589d6b0bd27212d/metrics.json).
## R01 — web 1440×900, dark
The scripted actions and actual video times below apply to the recording pair immediately following. Times are taken from each trace and corrected for the first CDP frame; action marks occur just before the interaction.
| Action | Before video seconds | After video seconds |
|---|---:|---:|
| toolbar open | 0.464 | 0.461 |
| toolbar close | 2.946 | 2.949 |
| Ctrl+J open | 4.105 | 4.111 |
| Ctrl+J close | 5.345 | 5.372 |
| toolbar open before drag | 6.429 | 6.603 |
| drag taller | 7.526 | 8.230 |
| drag shorter | 9.179 | 10.141 |
| close before reversal | 10.821 | 11.834 |
| rapid open | 11.880 | 12.894 |
| rapid close | 11.992 | 12.982 |
| rapid reopen | 12.078 | 13.048 |
**Before:** [R01-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0805a7550ee21ac7/R01-before.mp4) — [20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718). **After:** [R01-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0a1dd691a1a01acf/R01-after.mp4) — [f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a).
Video review frames: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/fda654a0688a4cee/R01-before-review.png) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a0f28d909e45da70/R01-after-review.png). Resize paint contacts: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/04f508b233ade266/R01-before-resize-paints.png) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ccecc553bd8f6167/R01-after-resize-paints.png).
Distance series CSV: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/eeee970b9061e020/R01-before-RO-series.csv) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9417b925e3dccec8/R01-after-RO-series.csv). Complete traces: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/43d544880073c09c/R01-before-trace.json) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c36f7f81dae8bb71/R01-after-trace.json).
Quiet working turn, held at edit approval: content 4048 px, no anchor, initial physical gap 0, following=true. Toolbar, keybinding, upward and downward handle drags, and a rapid open/close/reopen sequence all appear. The actual reversal intervals are computed from the timestamps above and remain below the 200 ms panel animation. The first handle drag is upward from the tall drawer's boundary toward y=200; the second is downward toward y=550. The corrected accepted driver therefore exercises both directions.
**Exact painted-height transient pair:** before record 645.2ms and after 697.5ms, painted viewport 384px on both sides. Nearest rAF gaps 551 / 12px. This is the largest before sampled-gap frame among the shared recorded painted heights in the selected action window; it is not a settled screenshot. At this phase the before content slips under the composer. The difference is app behavior, not a stale overlay or capture ghost.
| Before | After | Diff0 (amplified16×) | DiffT (>16) |
|---|---|---|---|
| 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718) | 
[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) |
Sheet: [R01-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/3e3dcdea7c612b10/R01-sheet.png). Full-frame changed pixels: Diff0 117571; DiffT 38748.
**Worst recorded sampled-gap view during this resize window:** before645.2ms, gap551, painted height384.0px; nearest available after painted height384.0px at697.5ms, gap12. Height mismatch 0px. The RO maximum may occur between painted frames; the sampled gap and pixel phase need not be simultaneous because rAF/RO/compositor run at different points. The exact-height pair above remains the primary image comparison.
| Before | After | Diff0 | DiffT (>16) |
|---|---|---|---|
| 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718) | 
[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) |
Supplementary sheet: [R01-worst-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0776c713243444fc/R01-worst-sheet.png). Selection metadata: [R01-phase-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/51dfe5aa0b9b79a0/R01-phase-selection.json) · [R01-worst-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b1aa82c797dd0864/R01-worst-selection.json).
## R02 — web 1440×900, dark
The scripted actions and actual video times below apply to the recording pair immediately following. Times are taken from each trace and corrected for the first CDP frame; action marks occur just before the interaction.
| Action | Before video seconds | After video seconds |
|---|---:|---:|
| approve: start extended synthetic Replay reply | 0.481 | 0.483 |
| close tall drawer during streaming | 2.814 | 1.766 |
| reopen tall drawer during streaming | 3.347 | 2.313 |
| drawer settled; observe continued streaming | 4.107 | 2.789 |
| post-resize continuation complete | 7.712 | 6.339 |
**Before:** [R02-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/8c83c269dd31b50f/R02-before.mp4) — [20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718). **After:** [R02-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/40cf240a68959d45/R02-after.mp4) — [f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a).
Video review frames: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7ecf08aa9b74090b/R02-before-review.png) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/875132f6a02edd26/R02-after-review.png). Resize paint contacts: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/df9859669836d879/R02-before-resize-paints.png) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5a5f8c749ab227d5/R02-after-resize-paints.png).
Distance series CSV: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/2976321ee1c4fc20/R02-before-RO-series.csv) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b21e75bae67230ff/R02-after-RO-series.csv). Complete traces: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/df96e7cb08ad11e6/R02-before-trace.json) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e3be7f19b770dd77/R02-after-trace.json).
The closed viewport848px shrinks to246px when the tall drawer reopens, a602px change larger than the new viewport. Markdown continues during the resize and long after it settles. In the after trace the text count grows on the same height-changing observer delivery (737→914chars), and that delivery reports gap0.
| Post-settle continuation | Before | After |
|---|---:|---:|
| Visible assistant chars | 1268→3392 | 914→3392 |
| Content height px | 4435→5161 | 4379→5161 |
| Following in every sampled continuation frame | True | True |
Example after content-only smooth-follow wave (constant viewport; record clock, not video clock):
| Record ms | Content height | Scroll top | Gap | Visible chars |
|---:|---:|---:|---:|---:|
| 2974.2 | 4491 | 4133 | 112 | 1268 |
| 3057.5 | 4491 | 4215 | 30 | 1268 |
| 3074.2 | 4491 | 4227 | 18 | 1268 |
| 3140.8 | 4491 | 4244 | 1 | 1268 |
| 3407.5 | 4602 | 4245 | 111 | 1622 |
**Exact painted-height transient pair:** before record3926.1ms and after2615.7ms, painted viewport246px on both sides. Nearest rAF gaps211 / 0px. This is the largest before sampled-gap frame among the shared recorded painted heights in the selected action window; it is not a settled screenshot. At this phase the before content slips under the composer. The difference is app behavior, not a stale overlay or capture ghost.
| Before | After | Diff0 (amplified16×) | DiffT (>16) |
|---|---|---|---|
| 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718) | 
[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) |
Sheet: [R02-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e1fcce635a61d9f2/R02-sheet.png). Full-frame changed pixels: Diff0 48687; DiffT 19581.
**Worst recorded sampled-gap view during this resize window:** before3645.9ms, gap455, painted height343.0px; nearest available after painted height310.0px at2585.4ms, gap0. Height mismatch33px. The RO maximum may occur between painted frames; the sampled gap and pixel phase need not be simultaneous because rAF/RO/compositor run at different points. The exact-height pair above remains the primary image comparison.
| Before | After | Diff0 | DiffT (>16) |
|---|---|---|---|
| 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718) | 
[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) |
Supplementary sheet: [R02-worst-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/84f1b8b22834110a/R02-worst-sheet.png). Selection metadata: [R02-phase-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a8970e6b4b8a042d/R02-phase-selection.json) · [R02-worst-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/1bbcb9873540658b/R02-worst-selection.json).
## R07b — web 1440×900, dark
The scripted actions and actual video times below apply to the recording pair immediately following. Times are taken from each trace and corrected for the first CDP frame; action marks occur just before the interaction.
| Action | Before video seconds | After video seconds |
|---|---:|---:|
| open drawer during working threshold crossing | 0.471 | 0.480 |
| approve and start assistant streaming while working | 0.898 | 0.926 |
| close drawer during continued working stream | 1.831 | 2.133 |
| reopen drawer during continued working stream | 2.345 | 2.642 |
| drawer settled; continued working text grows | 2.828 | 3.082 |
| working continuation complete | 6.337 | 6.598 |
**Before:** [R07b-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e773b86fa9a60974/R07b-before.mp4) — [20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718). **After:** [R07b-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/bcb1b773e6d5dbb1/R07b-after.mp4) — [f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a).
Video review frames: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/51b5ca70f2cd3e21/R07b-before-review.png) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/cc9844b86265e8bd/R07b-after-review.png). Resize paint contacts: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/3268c42779f7e099/R07b-before-resize-paints.png) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d1d390dacbcbe7be/R07b-after-resize-paints.png).
Distance series CSV: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/cca8e1d40f76cd75/R07b-before-RO-series.csv) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/8ce8d79c14f5fb2d/R07b-after-RO-series.csv). Complete traces: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/26c5954e2bcf31c0/R07b-before-trace.json) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e8d79d4afe8302f5/R07b-after-trace.json).
Initial content762px fits the848px closed viewport. Opening the180px drawer reduces the viewport to668px and creates94px overflow during a working approval. The script then approves and closes/reopens the drawer during real markdown continuation, remaining working throughout. This is the working boundary case, not an idle first-overflow capture.
| Post-settle continuation | Before | After |
|---|---:|---:|
| Visible assistant chars | 914→3215 | 1091→3215 |
| Content height px | 1093→1819 | 1149→1819 |
| Following in every sampled continuation frame | True | True |
Example after content-only smooth-follow wave (constant viewport; record clock, not video clock):
| Record ms | Content height | Scroll top | Gap | Visible chars |
|---:|---:|---:|---:|---:|
| 3460.8 | 1205 | 481 | 56 | 1268 |
| 3594.1 | 1205 | 524 | 13 | 1268 |
| 3627.5 | 1205 | 534 | 3 | 1268 |
| 3794.2 | 1205 | 537 | 0 | 1268 |
| 4444.2 | 1372 | 691 | 13 | 1799 |
**Exact painted-height transient pair:** before record762.3ms and after833.4ms, painted viewport680px on both sides. Nearest rAF gaps28 / 0px. This is the largest before sampled-gap frame among the shared recorded painted heights in the selected action window; it is not a settled screenshot. At this phase the before content slips under the composer. The difference is app behavior, not a stale overlay or capture ghost.
| Before | After | Diff0 (amplified16×) | DiffT (>16) |
|---|---|---|---|
| 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718) | 
[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) |
Sheet: [R07b-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/4869909ad6dbb6a9/R07b-sheet.png). Full-frame changed pixels: Diff0 101715; DiffT 83305.
**Worst recorded sampled-gap view during this resize window:** before2765.6ms, gap67, painted height671.0px; nearest available after painted height669.0px at3084.4ms, gap0. Height mismatch2px. The RO maximum may occur between painted frames; the sampled gap and pixel phase need not be simultaneous because rAF/RO/compositor run at different points. The exact-height pair above remains the primary image comparison.
| Before | After | Diff0 | DiffT (>16) |
|---|---|---|---|
| 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718) | 
[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) |
Supplementary sheet: [R07b-worst-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/22d0a7cc38ec1347/R07b-worst-sheet.png). Selection metadata: [R07b-phase-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/16a44c97222d7cdc/R07b-phase-selection.json) · [R07b-worst-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0af4b649f5327843/R07b-worst-selection.json).
## R09b — web 1440×900, dark
The scripted actions and actual video times below apply to the recording pair immediately following. Times are taken from each trace and corrected for the first CDP frame; action marks occur just before the interaction.
| Action | Before video seconds | After video seconds |
|---|---:|---:|
| release follow: wheel up while streaming | 0.475 | 0.484 |
| press actual end button | 1.224 | 1.236 |
| physical end reached; close drawer | 2.111 | 2.158 |
| reopen after settled re-follow | 2.703 | 2.641 |
| streaming continues after settled re-follow resize | 3.214 | 3.132 |
| overlap: release again | 5.421 | 5.336 |
| overlap: press end | 6.030 | 5.972 |
| overlap: close drawer before return settles | 6.151 | 6.094 |
| overlap: reopen drawer | 6.725 | 6.609 |
| streaming continues after overlapping resize | 7.294 | 7.157 |
**Before:** [R09b-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/6028a5add3f55fd6/R09b-before.mp4) — [20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718). **After:** [R09b-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d84f098eee822ecc/R09b-after.mp4) — [f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a).
Video review frames: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c97af961825c6301/R09b-before-review.png) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ee8154cad1db5079/R09b-after-review.png). Resize paint contacts: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/13004e3b3377f1b1/R09b-before-resize-paints.png) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7b5656d45f10ac3f/R09b-after-resize-paints.png).
Distance series CSV: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d120f53cd1e2e33b/R09b-before-RO-series.csv) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/8f5124630fd74970/R09b-after-RO-series.csv). Complete traces: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5734e37cb8dc3095/R09b-before-trace.json) · [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/430674ddb6cde416/R09b-after-trace.json).
The first wheel action releases following while markdown continues. The actual end button is clicked, then the driver waits until physical distance<=1 before resizing. The after physical-end milestone is gap1, following=true, at2.158s; before is gap1 at2.111s. The second variant resizes121ms after the end click while the return is still in progress. The after gap at that close action is748px, so it really overlaps the return; it is not another settled-end test.
Settled-return resize height-change peak: **591→0px**. Overlapping-return resize height-change peak: **643→0px**. Following is deliberately false while reading history; that is expected. The after recording continues from5162 to12950 visible assistant characters, finishes following=true, and retains normal smooth content-follow lag (final sampled gap29px).
This proves the requested user flows. It does not directly isolate the internal following=false plus<=40px branch, because following is already true at the physical-end milestone. No exact branch-coverage claim is made.
**Exact painted-height transient pair:** before record2930.1ms and after2913.7ms, painted viewport343px on both sides. Nearest rAF gaps591 / 0px. This is the largest before sampled-gap frame among the shared recorded painted heights in the selected action window; it is not a settled screenshot. At this phase the before content slips under the composer. The difference is app behavior, not a stale overlay or capture ghost.
| Before | After | Diff0 (amplified16×) | DiffT (>16) |
|---|---|---|---|
| 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718) | 
[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) |
Sheet: [R09b-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/2a663d49d386e3ef/R09b-sheet.png). Full-frame changed pixels: Diff0 114525; DiffT 35947.
**Worst recorded sampled-gap view during this resize window:** before2930.1ms, gap591, painted height343.0px; nearest available after painted height343.0px at2913.7ms, gap0. Height mismatch 0px. The RO maximum may occur between painted frames; the sampled gap and pixel phase need not be simultaneous because rAF/RO/compositor run at different points. The exact-height pair above remains the primary image comparison.
| Before | After | Diff0 | DiffT (>16) |
|---|---|---|---|
| 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718) | 
[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) | 
[20c3dd3689](https://github.com/pingdotgg/t3code/commit/20c3dd3689782bfcb1a04332086d08017028c718)→[f57ff765d8](https://github.com/pingdotgg/t3code/commit/f57ff765d849ae3b730e7ccdc5bbd86a0219c97a) |
Supplementary sheet: [R09b-worst-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7fe9d8be50ef883d/R09b-worst-sheet.png). Selection metadata: [R09b-phase-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/90aabc6d79e8f5c7/R09b-phase-selection.json) · [R09b-worst-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/effd6b4242969ff3/R09b-worst-selection.json).
## Settled control and inspection
Quiet R01 control captures were taken500ms apart and are pixel-identical within each side. Both gap0, working=true. The corrected R01 cross-side controls are also pixel-identical (Diff0 and DiffT both 0 pixels). Transient stills above are explicitly phase-matched, not presented as settled controls.
| Before first | Before +500ms | After first | After +500ms |
|---|---|---|---|
|  |  |  |  |
[R01-settled-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/eb3d9d34dc9946bf/R01-settled-sheet.png) · [R01-settled-Diff0.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/cd28a25d15769c53/R01-settled-Diff0.png) · [R01-settled-DiffT.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a336e7d96a97a0fc/R01-settled-DiffT.png).
Accepted before/after stills, difference images, resize-paint contacts, charts, settled controls, and encoded-video review contacts were opened and visually inspected. Videos were decoded into review frames. Normal working/Thinking indicators belong to the synthetic turn and are intentional. Before text covered by the composer is the defect under test; it was preserved. No loading overlay or blank rendering frame was accepted.
## Diff method and what else changed
Diffs are unmasked over the entire1440×900 frame. Per pixel, max absolute RGB channel delta is used. Diff0 highlights every nonzero delta and amplifies intensity16×. DiffT uses strictly>16 on8bit channels. Magenta/red highlights are composited over a20% brightness before image.
- Synthetic background-thread sidebar statuses, order and counters differ between sides because earlier fixture turns completed at different times. Those pixels remain in the full-frame diffs; no hidden mask removes them.
- Stream arrival timing differs, so transient paragraph numbers and character counts can differ despite identical payloads and interaction scripts. Compare timeline geometry, observer delivery and continuation, not exact prose at one wall-clock instant.
- The exact-height primary pairs share the painted drawer phase. Their record times differ. Worst-frame supplementary pairs disclose the available height mismatch.
- Compositor/subpixel anti-aliasing and cursor/edge pixels contribute small Diff0 changes.
- Clipped before text is the app defect. The working state, red Stop button and Thinking indicator are expected.
- The corrected R01 active-thread control matches pixel-for-pixel across sides; streaming rows still include the sidebar timing differences listed above. No application implementation was changed beyond the two supplied commits.
## Reproduction and complete source record
[Reproduction.md](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/761e16dce8ef4811/Reproduction.md) · [capture-r2-scripts.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/98b29aaaca0d0108/capture-r2-scripts.zip) · [capture-r2-metadata.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/82e0c1f39b38f1dd/capture-r2-metadata.zip).
The script bundle includes every seed/driver/analysis/helper script written for this step, the exact long and short Replay payloads, and the three reused prior-driver dependencies. It also contains abandoned helpers for provenance; accepted run paths are specified in metrics.json. Metadata includes all raw-source archive manifests, trace/results, inspection composites and redacted runtime logs. Credential-bearing/model-metadata log lines are removed; private auth/storage files are excluded.
[archive-verification.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/2279aefd9b9c5658/archive-verification.json) confirms decoded RGB equality for every original PNG frame in the lossless frame-index archives. Those MKVs preserve source pixels and original frame order, including rejected exploratory captures. They are not the timed proof videos; use their metadata timestamps where available. Rejected aborted runs without timestamps cannot reconstruct timing. Only source paths in metrics.json support the conclusions above.
[make-fixture.py](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c538e72b9c59a1b1/make-fixture.py)
[seed-clean-long.py](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/947997b40ef5fd47/seed-clean-long.py)
[seed-short.py](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/3f2f115c9613becc/seed-short.py)
[capture-r01.js](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f67ae98232cc1c68/capture-r01.js)
[capture-stream-final.js](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b7758eb92d78f107/capture-stream-final.js)
[capture-short.js](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c295c8c346b85c5f/capture-short.js)
[install-current.js](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0d9e46447558df97/install-current.js)
[cleanup-verification.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/cacde64f63a78f84/cleanup-verification.json) records that both capture worktrees were removed and every captured process leader was absent.
[public-url-verification.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b833eeb85d975153/public-url-verification.json) records successful public GET checks for the report, four sheets, eight timed videos and the reproduction bundles.
## Scope limits
All four requested1440-wide web rows are captured. Optional390-wide R01 was not captured. This is web-only evidence; no Electron shell or native mobile pass is claimed. Other matrix rows from previous rounds were not repeated. The particular internal<=40px following=false branch was not directly isolated. No real paid provider turn was used. No repository check suite was run because no implementation/test changes were made by this evidence step.
## Upload output (verbatim)
The following is ghdrop upload -m stdout, preserved verbatim. The linked evidence in the row sections is authoritative: upload-corrected supersedes earlier R01 versions in upload-public, and final bundles supersede upload-initial-bundles. Raw archives retain rejected runs for provenance. Video links are ordinary links, never image syntax.
upload-bundles.md
[capture-r2-scripts.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/781da59c8b240084/capture-r2-scripts.zip)
[capture-r2-metadata.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d4aa92a346f4f5ef/capture-r2-metadata.zip)
[Reproduction.md](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/761e16dce8ef4811/Reproduction.md)
[make-fixture.py](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c538e72b9c59a1b1/make-fixture.py)
[seed-clean-long.py](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/947997b40ef5fd47/seed-clean-long.py)
[seed-short.py](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/3f2f115c9613becc/seed-short.py)
[capture-r01.js](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f67ae98232cc1c68/capture-r01.js)
[capture-stream-final.js](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b7758eb92d78f107/capture-stream-final.js)
[capture-short.js](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c295c8c346b85c5f/capture-short.js)
[install-current.js](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0d9e46447558df97/install-current.js)
upload-corrected.md
[R01-after-RO-series.csv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9417b925e3dccec8/R01-after-RO-series.csv)
[R01-after-frames.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a9165d2126b39e4a/R01-after-frames.json)
[R01-after-trace.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c36f7f81dae8bb71/R01-after-trace.json)
[R01-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0a1dd691a1a01acf/R01-after.mp4)
[R01-before-RO-series.csv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/eeee970b9061e020/R01-before-RO-series.csv)
[R01-before-frames.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c5f7d52994454726/R01-before-frames.json)
[R01-before-trace.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/43d544880073c09c/R01-before-trace.json)
[R01-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0805a7550ee21ac7/R01-before.mp4)
[R01-phase-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/51dfe5aa0b9b79a0/R01-phase-selection.json)
[R01-worst-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b1aa82c797dd0864/R01-worst-selection.json)
[metrics.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/3589d6b0bd27212d/metrics.json)
[diff-statistics.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e8fe71b286e0a7aa/diff-statistics.json)
[archive-verification.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/2279aefd9b9c5658/archive-verification.json)
[run-2026-11-10T16-48-36.595Z-before-R01.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f361d975571cd5b3/run-2026-11-10T16-48-36.595Z-before-R01.mkv)
[run-2026-11-10T16-49-03.256Z-after-R01.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/688b93fc39ded6de/run-2026-11-10T16-49-03.256Z-after-R01.mkv)
upload-final-bundles.md
[capture-r2-scripts.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c123adf3cfd8d539/capture-r2-scripts.zip)
[capture-r2-metadata.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/1f3f2835e9c88fd4/capture-r2-metadata.zip)
[cleanup-verification.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/cacde64f63a78f84/cleanup-verification.json)
upload-initial-bundles.md
[capture-r2-scripts.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/4d8ac7d8dbe9c266/capture-r2-scripts.zip)
[capture-r2-metadata.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d3e223fcb572a335/capture-r2-metadata.zip)
upload-public.md
[R01-after-RO-series.csv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a20cd0d7aef9935b/R01-after-RO-series.csv)
[R01-after-frames.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/32501ea47d9ac1d3/R01-after-frames.json)
[R01-after-trace.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9784f2ce51746582/R01-after-trace.json)
[R01-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/6b067c2fe517b06d/R01-after.mp4)
[R01-before-RO-series.csv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ad91154d5fbecb73/R01-before-RO-series.csv)
[R01-before-frames.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/23192cde7da9475b/R01-before-frames.json)
[R01-before-trace.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/83014226c1bc4bdc/R01-before-trace.json)
[R01-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/92d4b29fbed281c9/R01-before.mp4)
[R01-phase-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ce80d1fe757d89a9/R01-phase-selection.json)
[R01-worst-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/fa026b1e85e4500f/R01-worst-selection.json)
[R02-after-RO-series.csv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b21e75bae67230ff/R02-after-RO-series.csv)
[R02-after-frames.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/cc3c465bdc1a28ce/R02-after-frames.json)
[R02-after-trace.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e3be7f19b770dd77/R02-after-trace.json)
[R02-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/40cf240a68959d45/R02-after.mp4)
[R02-before-RO-series.csv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/2976321ee1c4fc20/R02-before-RO-series.csv)
[R02-before-frames.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ea20e9b18da1e739/R02-before-frames.json)
[R02-before-trace.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/df96e7cb08ad11e6/R02-before-trace.json)
[R02-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/8c83c269dd31b50f/R02-before.mp4)
[R02-phase-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a8970e6b4b8a042d/R02-phase-selection.json)
[R02-worst-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/1bbcb9873540658b/R02-worst-selection.json)
[R07b-after-RO-series.csv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/8ce8d79c14f5fb2d/R07b-after-RO-series.csv)
[R07b-after-frames.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/3b96ccec413ed58a/R07b-after-frames.json)
[R07b-after-trace.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e8d79d4afe8302f5/R07b-after-trace.json)
[R07b-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/bcb1b773e6d5dbb1/R07b-after.mp4)
[R07b-before-RO-series.csv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/cca8e1d40f76cd75/R07b-before-RO-series.csv)
[R07b-before-frames.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/276c28ebb31e4bbc/R07b-before-frames.json)
[R07b-before-trace.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/26c5954e2bcf31c0/R07b-before-trace.json)
[R07b-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e773b86fa9a60974/R07b-before.mp4)
[R07b-phase-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/16a44c97222d7cdc/R07b-phase-selection.json)
[R07b-worst-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0af4b649f5327843/R07b-worst-selection.json)
[R09b-after-RO-series.csv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/8f5124630fd74970/R09b-after-RO-series.csv)
[R09b-after-frames.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9c369a47f1c25636/R09b-after-frames.json)
[R09b-after-trace.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/430674ddb6cde416/R09b-after-trace.json)
[R09b-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d84f098eee822ecc/R09b-after.mp4)
[R09b-before-RO-series.csv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d120f53cd1e2e33b/R09b-before-RO-series.csv)
[R09b-before-frames.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a6da2586059bf0f1/R09b-before-frames.json)
[R09b-before-trace.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5734e37cb8dc3095/R09b-before-trace.json)
[R09b-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/6028a5add3f55fd6/R09b-before.mp4)
[R09b-phase-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/90aabc6d79e8f5c7/R09b-phase-selection.json)
[R09b-worst-selection.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/effd6b4242969ff3/R09b-worst-selection.json)
[diff-statistics.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c391f6c5c779e9e0/diff-statistics.json)
[metrics.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/459d4383d8b87f11/metrics.json)
upload-raw.md
[run-2026-11-10T16-13-01.823Z-before-R01.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/dbf823291bbc65e3/run-2026-11-10T16-13-01.823Z-before-R01.mkv)
[run-2026-11-10T16-13-01.868Z-after-R01.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/afbca5568367acc7/run-2026-11-10T16-13-01.868Z-after-R01.mkv)
[run-2026-11-10T16-14-06.327Z-before-R01.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0e3dbb7c24ec854a/run-2026-11-10T16-14-06.327Z-before-R01.mkv)
[run-2026-11-10T16-14-06.331Z-after-R01.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/378cd900c1a4da50/run-2026-11-10T16-14-06.331Z-after-R01.mkv)
[run-2026-11-10T16-14-44.969Z-after-R02.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/bf06076f7d2204a6/run-2026-11-10T16-14-44.969Z-after-R02.mkv)
[run-2026-11-10T16-14-45.161Z-before-R02.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/3d828faf13961680/run-2026-11-10T16-14-45.161Z-before-R02.mkv)
[run-2026-11-10T16-17-07.451Z-after-R01.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/30c859cd05dd1193/run-2026-11-10T16-17-07.451Z-after-R01.mkv)
[run-2026-11-10T16-17-07.483Z-before-R01.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/008a10ab62716a99/run-2026-11-10T16-17-07.483Z-before-R01.mkv)
[run-2026-11-10T16-17-47.595Z-after-R02.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ac8d644d2f95e61e/run-2026-11-10T16-17-47.595Z-after-R02.mkv)
[run-2026-11-10T16-17-47.625Z-before-R02.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f820d6c815df8fb9/run-2026-11-10T16-17-47.625Z-before-R02.mkv)
[run-2026-11-10T16-21-10.797Z-after-R02.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/97c2ec193182eaa7/run-2026-11-10T16-21-10.797Z-after-R02.mkv)
[run-2026-11-10T16-21-10.797Z-after-R09b.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0c94fc13dd33386e/run-2026-11-10T16-21-10.797Z-after-R09b.mkv)
[run-2026-11-10T16-21-10.829Z-before-R02.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a732aaecd46f2163/run-2026-11-10T16-21-10.829Z-before-R02.mkv)
[run-2026-11-10T16-21-10.829Z-before-R09b.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/669563efe9c50377/run-2026-11-10T16-21-10.829Z-before-R09b.mkv)
[run-2026-11-10T16-25-36.579Z-before-R07b.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/dd925dbde60f1722/run-2026-11-10T16-25-36.579Z-before-R07b.mkv)
[run-2026-11-10T16-26-40.867Z-after-R07b.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/512b116b63dac2cb/run-2026-11-10T16-26-40.867Z-after-R07b.mkv)
[run-2026-11-10T16-29-07.370Z-before-R01.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f7ef019c80f342f2/run-2026-11-10T16-29-07.370Z-before-R01.mkv)
[run-2026-11-10T16-30-03.708Z-after-R01.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/8c18bf0b661fc13a/run-2026-11-10T16-30-03.708Z-after-R01.mkv)
[run-2026-11-10T16-30-42.050Z-before-R02.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c00eeacfd1c04b82/run-2026-11-10T16-30-42.050Z-before-R02.mkv)
[run-2026-11-10T16-30-42.050Z-before-R09b.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/eb22b55fe10e48ad/run-2026-11-10T16-30-42.050Z-before-R09b.mkv)
[run-2026-11-10T16-32-14.838Z-after-R02.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ab261c19056b25ec/run-2026-11-10T16-32-14.838Z-after-R02.mkv)
[run-2026-11-10T16-32-14.838Z-after-R09b.mkv](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/1de19f821b545e64/run-2026-11-10T16-32-14.838Z-after-R09b.mkv)
upload-verified-bundles.md
[capture-r2-scripts.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/98b29aaaca0d0108/capture-r2-scripts.zip)
[capture-r2-metadata.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/82e0c1f39b38f1dd/capture-r2-metadata.zip)
[public-url-verification.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b833eeb85d975153/public-url-verification.json)
AI-assisted; verified with the tests and tools listed above.
FRICTION:
- tool | An initial HTTP HEAD probe returned403, while browser-style public GET requests returned200 | Validate downloadable artifacts with GET.
- tool | Shared preview click failed twice ("Provide exactly one target"; "received an invalid locator"); agent-browser absent | A working preview locator or installed CLI would have avoided the Playwright fallback.
- runbook | Timeline fiber message props stayed stale while live message text grew | Sample visible DOM text and list state separately.
- infra | Rejected recording attempts left active screencast sessions and slowed capture | Stop/detach sessions in finally; a fresh browser removed the interference.
- model | Initial quiet-turn drag targets both shortened the drawer; caught during final inspection and recaptured in both directions | Verify target direction against the actual starting drawer boundary.
- model | Earlier driver iteration used a wrong live entry index and missing null guard | Inspect current root and guard unmounted timeline reads.
Captured 4/4 requested rows; reproduction and source artifacts uploaded.