## Evidence
**Additional pixel variation — resolved by outcome (a):** R01 and R03 each have the same nine pixels at x=4..222, y=68..70, maximum channel delta 1/255. Fresh HEAD controls reproduce the BASE-before band exactly when Z is open (0/5,376 changed pixels in x=0..255, y=60..80), and reproduce the original AFTER band on the deletion destination (0/5,376 changed pixels). After deleting D and clicking Z on HEAD, both complete frames match the original BASE-before stills exactly (0/1,296,000 changed pixels, zero threshold). The nine pixels are therefore navigation-dependent rendering associated with the intended route transition, not a fixed build difference. Navigation history also matters: in R01, reopening A after Z retains the before-band and differs from the original AFTER only at those nine pixels. No claim that final thread identity alone determines rasterization is made. Original recordings and unmasked Diff0/DiffT are retained. R03 shows the existing draft UI because the destination changes, including its default model and permissions controls.
Captured against merge-base [30084f0018](https://github.com/pingdotgg/t3code/commit/30084f0018c5751a07291222e8d04e16d513700a) and head [ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) for [#29643](https://git.cafe/pingdotgg/t3code/pull/29643), [GitHub mirror #29643](https://github.com/pingdotgg/t3code/pull/29643). **7/7 required rows, 10/10 required recording runs; 7/15 total matrix rows.** All seven rows use the real app and synthetic RPC-created data. R05 uses a scratch scripted provider; no client harness or injected shell was needed.
Web environment (R01–R06): Playwright-driven headless Chromium 148.0.7778.96 on Linux; no device emulation; 1440×900 CSS px, DPR 1, browser zoom 1, font scaling 100% (root 16px), dark theme, reduced motion. Before and after use the same seed driver, thread titles/ranking, scroll origin and fixed browser/server clock at 2026-11-03T10:30:00.000Z. R04 explicitly advances both clocks to 10:30:11. Synthetic providers have completed before each capture except the deliberately pending R05 request.
Desktop environment (R07): real Electron 44.10.0 / Chromium 152.0.7977.169 on Linux, headless-desktop Xvfb, ffmpeg x11grab recording and ImageMagick screen capture. Exact content bounds x=0, y=0, width=1440, height=900 CSS px; DPR 1, zoom 1, font scaling 100% (root 16px), no device emulation, dark app, reduced motion. The native Linux menu uses its light system appearance. The browser clock is fixed at 10:30; the isolated bundled server uses its actual wall clock (timestamps are in the transcript).
Each initial/final still waits for network idle and fonts, then requires two pixel-identical captures 500 ms apart. Desktop stills use the same rule on raw Xvfb pixels, including the native menu still. Captures, comparison sheets and video contact sheets were opened and inspected. Contact sheets are derived video frames, not replacement still proof. All recordings show the initial route, actual menu/delete interaction, confirmation where enabled, and stable result. Narration times below are approximate driver seconds from the settled initial state; video encoding introduces a small subsecond offset.
Diff0 highlights **every nonzero maximum-channel difference** in bright magenta (including a one-level change). DiffT uses **maximum channel difference >16/255**. Unchanged areas are a dimmed after image. Differences in selected thread/route and their existing UI are intended.
Common ranking data (R01/R02, UTC): A created 10:01, latest user send 10:11; B created 10:02, send 10:10; Z created 10:03, send 10:12; D created 10:04, send 10:13. Thus Z outranks A/B under both sorts before eligibility filtering; A wins updated_at, B wins created_at. The seed/fixture transcripts contain actual shell values. All fixture threads disable auto-settle.
### R01 — Future snooze, two live survivors — web 1440×900, dark
Before opens Z, Parked candidate, despite its future 14:30 snooze. After opens A, Older created, recent send. Z stays in Snoozed (count 1), and B survives. Sort: updated_at. Header chevron and confirmation enabled.
**Same scripted interaction, before:** ~1.50s Click header title chevron; ~2.56s Choose Delete; ~3.68s Confirm Delete; ~5.75s Final settled route.
**After:** ~1.50s Click header title chevron; ~2.55s Choose Delete; ~3.63s Confirm Delete; ~6.12s Final settled route.
| Before | After | Diff0 | DiffT (>16/255) |
| --- | --- | --- | --- |
| [R01-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ce5f643b87fe8ee1/R01-before.mp4)
[30084f0018](https://github.com/pingdotgg/t3code/commit/30084f0018c5751a07291222e8d04e16d513700a) | [R01-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/37645dd1a96a3112/R01-after.mp4)
[ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) | 
[30084f0018](https://github.com/pingdotgg/t3code/commit/30084f0018c5751a07291222e8d04e16d513700a) → [ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) | 
[30084f0018](https://github.com/pingdotgg/t3code/commit/30084f0018c5751a07291222e8d04e16d513700a) → [ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) |
Comparison sheet: [R01-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9445125578782fd9/R01-sheet.png).
Before data: [R01-before-seed.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/1a91cd63757f6ac2/R01-before-seed.json), [R01-before-fixture.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/307d90f84c04dfdb/R01-before-fixture.json), [R01-before-transcript.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5ad6fb24cb485753/R01-before-transcript.json). After data: [R01-after-seed.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/02aa935c1352a63e/R01-after-seed.json), [R01-after-fixture.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/8078cef9890ca03f/R01-after-fixture.json), [R01-after-transcript.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/898bfb54b75c686f/R01-after-transcript.json).
Initial frames: [R01-before-initial.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ca33bbcf7040d9a3/R01-before-initial.png), [R01-after-initial.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/92e9facbefcd781a/R01-after-initial.png).
**Navigation/rendering control — outcome (a):** On HEAD [ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047), fresh generation `recapture2-r01-20261103` with the same seed, frozen clock and rendering tuple: directly opening Z before deleting anything matches the original BEFORE sidebar band at zero threshold (0 changed pixels; 9 versus AFTER, max 1/255). The full direct-Z frame differs from BEFORE by 4,402 pixels because D is still in the sidebar. After deleting D through the real header menu/confirmation, the native A result is pixel-identical to original AFTER (0/1,296,000). Clicking Z then gives a frame pixel-identical to original BASE-before (0/1,296,000). The same nine-pixel set is reproduced entirely within HEAD. Reopening A after Z differs from original AFTER by exactly those nine pixels and nothing else, so navigation history contributes to the rendering state. No base checkout/rebuild was necessary.
Raw HEAD-on-Z still: [R01-head-deleted-D-opened-Z.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/cc42538b00dd1ab9/R01-head-deleted-D-opened-Z.png); full-frame zero-threshold comparison: [R01-control-Diff0.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/eadd2c1014d1f0c2/R01-control-Diff0.png). Measurements: [control-statistics.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/be12b735c32fc627/control-statistics.json), exact pixel coordinates: [band-pixel-verification.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/98bb4d313d7063f4/band-pixel-verification.json). All raw controls, original reference stills, full-frame control diffs, seed/transcript data, inspected hashes, cleanup transcript and copied/adapted drivers: [control-artifacts.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/85c93f7796cf9171/control-artifacts.zip).
### R02 — Settled candidate, conflicting creation/send order — web 1440×900, dark
Before opens settled Z. After opens B, Newer created, older send. Z remains in Settled (count 1), A survives. Sort: created_at. Sidebar right-click and confirmation enabled.
**Same scripted interaction, before:** ~1.50s Right-click Delete me in sidebar; ~2.63s Choose Delete; ~3.73s Confirm Delete; ~6.66s Final settled route.
**After:** ~1.50s Right-click Delete me in sidebar; ~2.56s Choose Delete; ~3.64s Confirm Delete; ~6.42s Final settled route.
| Before | After | Diff0 | DiffT (>16/255) |
| --- | --- | --- | --- |
| [R02-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/bcc11d42f16dc1a3/R02-before.mp4)
[30084f0018](https://github.com/pingdotgg/t3code/commit/30084f0018c5751a07291222e8d04e16d513700a) | [R02-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/1708bf464ffd20fa/R02-after.mp4)
[ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) | 
[30084f0018](https://github.com/pingdotgg/t3code/commit/30084f0018c5751a07291222e8d04e16d513700a) → [ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) | 
[30084f0018](https://github.com/pingdotgg/t3code/commit/30084f0018c5751a07291222e8d04e16d513700a) → [ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) |
Comparison sheet: [R02-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/10eef2d3b0712009/R02-sheet.png).
Before data: [R02-before-seed.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a2b3ef3fcfb32bae/R02-before-seed.json), [R02-before-fixture.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f56a96457350a6f4/R02-before-fixture.json), [R02-before-transcript.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/37156b09522d2797/R02-before-transcript.json). After data: [R02-after-seed.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/046bf9a7b22ed529/R02-after-seed.json), [R02-after-fixture.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/abfc39936537f701/R02-after-fixture.json), [R02-after-transcript.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/dbbb2ff11d66aba1/R02-after-transcript.json).
Initial frames: [R02-before-initial.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/bde490cda64c7db3/R02-before-initial.png), [R02-after-initial.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7b710f749898cca2/R02-after-initial.png).
### R03 — All same-project survivors parked — web 1440×900, dark
Before opens snoozed Z. After routes to a thread draft in project P, Fallback fixture; it does not open Z, S or the eligible survivor Q in Other project. Both parked shelf counts remain 1. The completed draft landing is captured, including the existing hero and composer. Sort: updated_at. Header chevron and confirmation enabled.
**Same scripted interaction, before:** ~1.50s Click header title chevron; ~2.58s Choose Delete; ~3.67s Confirm Delete; ~5.79s Final settled route.
**After:** ~1.50s Click header title chevron; ~2.55s Choose Delete; ~3.63s Confirm Delete; ~6.41s Final settled route.
| Before | After | Diff0 | DiffT (>16/255) |
| --- | --- | --- | --- |
| [R03-before.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ff79e40cbc008374/R03-before.mp4)
[30084f0018](https://github.com/pingdotgg/t3code/commit/30084f0018c5751a07291222e8d04e16d513700a) | [R03-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ba06d5f03649036f/R03-after.mp4)
[ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) | 
[30084f0018](https://github.com/pingdotgg/t3code/commit/30084f0018c5751a07291222e8d04e16d513700a) → [ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) | 
[30084f0018](https://github.com/pingdotgg/t3code/commit/30084f0018c5751a07291222e8d04e16d513700a) → [ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) |
Comparison sheet: [R03-sheet.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/965bc87fa19341a8/R03-sheet.png).
Before data: [R03-before-seed.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b6d14744922e201f/R03-before-seed.json), [R03-before-fixture.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/41b3bcfa7de4388c/R03-before-fixture.json), [R03-before-transcript.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9d2b8ba266e4b319/R03-before-transcript.json). After data: [R03-after-seed.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0b1e3680510844f3/R03-after-seed.json), [R03-after-fixture.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0736f2cd0cd3ad66/R03-after-fixture.json), [R03-after-transcript.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0e8af5341a5ace38/R03-after-transcript.json).
Initial frames: [R03-before-initial.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9332eff55ce25589/R03-before-initial.png), [R03-after-initial.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9adfe41c1691ebdd/R03-after-initial.png).
**Navigation/rendering control — outcome (a):** On HEAD [ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047), fresh generation `recapture2-r03-20261103`: directly opening Z without deleting D matches original BEFORE's sidebar band exactly (0/5,376 changed pixels; 9 versus AFTER, max 1/255). The direct-Z full frame differs by 3,370 pixels because D still exists and shifts the active rows. Deleting D reaches the draft and matches the original AFTER sidebar band exactly (0 changed pixels). Clicking Z afterward matches the entire original BASE-before frame exactly (0/1,296,000). The same nine-pixel set is reproduced entirely within HEAD. The new draft control is not substituted for the original AFTER: it adds the existing “or start without a project” link and shifts its hero heading, producing 10,854 differing pixels at x=472..1023, y=316..373, while all other pixels, including the sidebar band, match original AFTER. Original stills, recordings and Diff0/DiffT are retained.
Raw HEAD-on-Z still: [R03-head-deleted-D-opened-Z.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9fc318b30501aa00/R03-head-deleted-D-opened-Z.png); full-frame zero-threshold comparison: [R03-control-Diff0.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f2534c119f21322e/R03-control-Diff0.png). Measurements: [control-statistics.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/be12b735c32fc627/control-statistics.json), exact pixel coordinates: [band-pixel-verification.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/98bb4d313d7063f4/band-pixel-verification.json). All raw controls, original reference stills, full-frame control diffs, seed/transcript data, inspected hashes, cleanup transcript and copied/adapted drivers: [control-artifacts.zip](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/85c93f7796cf9171/control-artifacts.zip).
### R04 — Timer wake is eligible again — web 1440×900, dark
At ~1.50s the driver advances browser/server time from 10:30:00 to 10:30:11, crossing Z’s 10:30:10 deadline. Z visibly moves out of Snoozed before deletion. Its raw snoozedUntil remains 10:30:10; deletion at 10:30:11 opens Z, not A. This is a controlled clock transition, not an eleven-second real-time wait.
**Recorded steps:** ~1.50s Advance server/browser clock past recorded snooze deadline; ~7.35s Observed Z awake in Active; raw snooze deadline retained; ~8.38s Click header title chevron; ~9.48s Choose Delete; ~10.56s Confirm Delete; ~13.22s Final settled route.
| Before | After | Diff0 | DiffT |
| --- | --- | --- | --- |
| Not required for this head control | [R04-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/71d8d0102588d070/R04-after.mp4)
[ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) | Not applicable: no before pair | Not applicable: no before pair |
Data: [R04-after-seed.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/20b31fc6a200ffd8/R04-after-seed.json), [R04-after-fixture.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a6fbe6254aa6efef/R04-after-fixture.json), [R04-after-transcript.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/4fa5687010422099/R04-after-transcript.json). Initial frame: [R04-after-initial.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/371f09b9d077a27a/R04-after-initial.png). Contact sheet: [R04-after-filmstrip.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e5e1e8997d00b3a8/R04-after-filmstrip.png).
Wake shell: [R04-after-wake-shell.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/81724630e1c105e7/R04-after-wake-shell.json).
### R06 — Bulk deletion, parked and pinned survivors — web 1440×900, dark
Confirmation is deliberately disabled in persisted client settings (see seed transcript). Control-click C then D, context Delete (2). Actual outgoing thread.delete commands are C at ~3.981s, then open D at ~4.199s. Final route opens pinned A; snoozed Z and settled S remain. The recorded route log contains no intermediate parked selection. Working shelf configuration is recorded; no worktrees are involved.
**Recorded steps:** ~1.50s Select Delete first, then open Delete me, with Control-click; ~2.82s Right-click selection; ~3.87s Choose Delete; ~6.73s Final settled route.
| Before | After | Diff0 | DiffT |
| --- | --- | --- | --- |
| Not required for this head control | [R06-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/93b95733a78bf6e0/R06-after.mp4)
[ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) | Not applicable: no before pair | Not applicable: no before pair |
Data: [R06-after-seed.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7c99ccdf4dabab34/R06-after-seed.json), [R06-after-fixture.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/663e4f52a07a7889/R06-after-fixture.json), [R06-after-transcript.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/bebe42176f407c34/R06-after-transcript.json). Initial frame: [R06-after-initial.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c57913b7665f1e49/R06-after-initial.png). Contact sheet: [R06-after-filmstrip.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/067096ecc8321123/R06-after-filmstrip.png).
### R07 — Native desktop menu — desktop 1440×900, dark
Real Electron header chevron opens the native Linux menu, captured at screen level. End, Return chooses its last entry, Delete, followed by the app’s Confirm dialog. A opens; Z remains Snoozed and B survives. The fixture uses the R01 ranking pattern and updated_at sort. Working shelf is disabled for this capture so A/B remain visible.
**Recorded steps:** ~1.50s Click header chevron to open native Electron menu; ~3.46s Choose native Delete with End, Return; ~4.54s Confirm app deletion dialog; ~6.07s Final settled A; Z remains Snoozed.
| Before | After | Diff0 | DiffT |
| --- | --- | --- | --- |
| Not required for this head control | [R07-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7023ec2d96828ea0/R07-after.mp4)
[ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) | Not applicable: no before pair | Not applicable: no before pair |
Data: [R07-after-seed.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e485de48db447e5f/R07-after-seed.json), [R07-after-fixture.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0dc424bcc7de057b/R07-after-fixture.json), [R07-after-transcript.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f546a53af088b258/R07-after-transcript.json). Initial frame: [R07-after-initial.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c14fa7ccbd68fead/R07-after-initial.png). Contact sheet: [R07-after-filmstrip.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a410fca3c274d7cd/R07-after-filmstrip.png).
Native menu still: [R07-after-native-menu.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5ff2cf852a1c283d/R07-after-native-menu.png).
### R05 — Early wake with future deadline — REAL APP / SCRIPTED PROVIDER — web 1440×900, dark
The scratch Fx provider waits on a release file before elicitation/create. Z is snoozed while its run is active, then the request is released. Persisted pendingRuntimeRequest.kind is user_input and blocking=true, while snoozedUntil remains 2026-11-03T14:30:00.000Z and now remains 10:30. Z has the most recent user send. Its Input state is visible before deleting D; deletion opens Z and displays the request choices. The final shell still contains the future deadline and outstanding request. No shell interception or client override is used.
**Recorded steps:** ~1.50s Click header title chevron; ~2.56s Choose Delete; ~3.65s Confirm Delete; ~6.55s Final settled route.
| Before | After | Diff0 | DiffT |
| --- | --- | --- | --- |
| Not required for this head control | [R05-after.mp4](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c29ca7f9016ebc25/R05-after.mp4)
[ace518f2a7](https://github.com/pingdotgg/t3code/commit/ace518f2a78ebd640e1d5b06a652bd9d5c42b047) | Not applicable: no before pair | Not applicable: no before pair |
Data: [R05-after-seed.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e8549aeacb54b126/R05-after-seed.json), [R05-after-fixture.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/1a0739a5523780d2/R05-after-fixture.json), [R05-after-transcript.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/1e8f007ccf0c3fad/R05-after-transcript.json). Initial frame: [R05-after-initial.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f059184a8ed31d78/R05-after-initial.png). Contact sheet: [R05-after-filmstrip.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/2a5acd92d90d6c20/R05-after-filmstrip.png).
Before releasing provider request: [R05-after-before-release-shell.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/471b936e17749160/R05-after-before-release-shell.json). After persisted request: [R05-after-after-release-shell.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e3b3d1da4c0f6485/R05-after-after-release-shell.json).
### What else changed
- R01: title and sidebar selection move from Z to A; Z loses its selected emphasis, A gains it; the snoozed banner disappears because A is open. Diff0 also contains the nine 1/255 pixels at the banner gradient’s lower edge (x=4..222, y=68..70). The HEAD controls above reproduce both bands, and HEAD-on-Z matches the original BASE-before full frame exactly. These pixels accompany the intended navigation transition; the controls rule out a fixed build difference and show that navigation history affects this rendering.
- R02: title and sidebar selection move from Z to B; Z loses selected emphasis, B gains it; the settled banner disappears. The shelf count and both eligible survivor identities remain.
- R03: title becomes New thread, Z loses selected emphasis, the prior user/assistant messages and snoozed banner disappear. The existing draft hero and centered composer appear, with its default 6 Astra / Full access controls instead of the thread’s eval/scripted / Supervised controls; Current checkout/branch controls appear below it. The right panel’s compact header and vertical placement differ for the draft route. Diff0 also shows the same nine 1/255 banner-gradient pixels; the HEAD controls reproduce the before and after bands, and HEAD-on-Z matches the original BASE-before full frame exactly. These draft render differences follow from the intended route change; the PR adds no JSX or styling.
- R04–R07 are head controls without a baseline comparison. Their intended visible transitions are narrated above; no before/after pixel claim is made for them.
### Coverage limits
**No required row is unverified.** R08a/R08b are unverified concurrent-client/worktree-confirmation risks and were explicitly excluded this round. C01–C06 were explicitly excluded optional controls/equivalence cases, so they have no recordings here. This evidence does not establish behavior for those eight rows, transport permutations, narrow/light variants or native mobile.
### Seed and driver scripts
All synthetic fixture and driver source stays outside the repository. Pairing tokens, auth storage, private server logs and real user data are not uploaded. The head checkout is unchanged; the PR body was not edited. Seed runs use fresh isolated databases with T3CODE_DEV_SEED=0; no live T3 home was copied. Reruns must use a new EVIDENCE_GENERATION because deleted IDs retain event-sourced tombstones.
- [start-app.mjs](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5e41bd5ef9a2996b/start-app.mjs)
- [clock-preload.mjs](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e8e1cd1fdde28217/clock-preload.mjs)
- [probe.mjs](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/472a0279471a84d8/probe.mjs)
- [browser-lib.mjs](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/87ec6de2590acd99/browser-lib.mjs)
- [seed.mjs](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b2a2db9e25df997b/seed.mjs)
- [scripted-provider.ts](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/510a71d6a6b7478e/scripted-provider.ts)
- [fx](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5e2da571003c9eff/fx)
- [capture.mjs](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9e39686368962a40/capture.mjs)
- [desktop-capture.mjs](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5e64f69dc950b39a/desktop-capture.mjs)
- [make-diffs.py](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/77c0139bd49c5a2c/make-diffs.py)
- [publish.py](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0001fe795b279360/publish.py)
- [evidence-index.py](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/16517f297fc61cb5/evidence-index.py)
- [rpc-probe.mjs](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/6c7b8c83e9407587/rpc-probe.mjs)
- [ui-probe.mjs](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/165bf9ccd36e684c/ui-probe.mjs)
- [diagnose.mjs](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a9eebd991c6d08ac/diagnose.mjs)
- [desktop-probe.mjs](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/79c9c296bc6f2773/desktop-probe.mjs)
- [finalize.py](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e5dd192fbb590ec0/finalize.py)
Cleanup: [cleanup-transcript.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f91df08c942c5462/cleanup-transcript.json). Both captured server process groups and their owned provider children are stopped, capture ports are closed, all captured Electron/recorder PIDs are gone, and the base worktree is removed. Head remains clean at the requested SHA. Disk availability after cleanup: 13 GB.
Complete verbatim ghdrop output: [ghdrop-output.md](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/028e7baef9fbe2b9/ghdrop-output.md).
Artifact manifest: [uploads.json](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/64892ce178a32dba/uploads.json).
Recapture cleanup: the captured HEAD server process group and its owned detached provider descendant are stopped; all three browser PIDs captured at launch are gone and the preview tab is closed. Ports 5919/13959 are closed. No base worktree was created. HEAD is clean at the requested SHA; disk availability is 13 GB. The bundle includes the recapture cleanup transcript. The original evidence directory and proof-audit-2 directory were read-only throughout. No worktree edits, commits, pushes, PR-body edits, new baseline captures, replacement recordings or masked diffs were made. The web controls cover both navigation destinations using the same Chromium 148.0.7778.96, 1440×900, DPR 1, dark theme, reduced motion, 16px root font, zoom 1 and 2026-11-03T10:30:00.000Z clock. Each still passed network-idle/fonts and two pixel-identical screenshots 500 ms apart.