Controlled-contention proof in the real web app (Chromium 156, Linux, 1280×800, light theme, fresh synthetic database per commit). Identical temporary lock-holder and controlled replay fixtures were applied to both revisions; run-end and queue-promotion code was untouched.

Create synthetic threads A and B and start controlled replay turns on both; queue a follow-up on B; open B and wait for its composer; acquire A's real ThreadCommandExecutor lock; finish A and wait for its persisted completed event, then finish B; capture B before releasing A's lock. Base keeps the follow-up queued; head starts it. Both sides use the same fixture and assertion.

### Scenario: An idle thread starts its queued follow-up while another thread's lock is busy

**Proven: the scenario fails on base [e367ef0da1](https://github.com/pingdotgg/t3code/commit/e367ef0da11e11a814911c93d812e2681c51294b) and passes on head [1235165531](https://github.com/pingdotgg/t3code/commit/1235165531e7abe49286016786a87697789eac5e).**

|  | Base [e367ef0da1](https://github.com/pingdotgg/t3code/commit/e367ef0da11e11a814911c93d812e2681c51294b) | Head [1235165531](https://github.com/pingdotgg/t3code/commit/1235165531e7abe49286016786a87697789eac5e) |
| --- | --- | --- |
| Result | ❌ failed<br>ASSERTION_FAILED: expect.toBeVisible failed<br>locator: getByText("Queued follow-up started while the unrelated busy lock is held.")<br>expected: visible<br>observed: no node (match count 0) | ✅ passed |
| stability check | ![Base e367ef0da1: stability check](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/8ff928c0d2365a3e/001-stability-check.png) | ![Head 1235165531: stability check](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7dc0d77826a2f660/001-stability-check.png) |
| idle queue with unrelated lock held | ![Base e367ef0da1: idle queue with unrelated lock held](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5d4e5c8e1a90949f/002-idle-queue-with-unrelated-lock-held.png) | ![Head 1235165531: idle queue with unrelated lock held](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f6183748414ae815/002-idle-queue-with-unrelated-lock-held.png) |
| failure | ![Base e367ef0da1: failure](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/1d71a67f91c82c9a/003-failure.png) |  |

| Recording | Before | After |
| --- | --- | --- |
| Same 9-second interaction | [Before recording](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/4986e5032526c7bd/before.mp4) | [After recording](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/36f7fecf596bfa69/after.mp4) |

[Scenario](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/57e4a55183836322/busy-thread-lock.e2e.ts) · [Lock/replay fixture](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e44cce5e9b1873c5/instrument.py) · [Startup hook](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/8de4170f8c807265/start-app-hook.txt) · [Reproduction guide](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e5956e404738e579/README.md) · [Runtime event audit](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/96be3f46a609fa0f/runtime-audit.json)

The audit records A completed at event 120, then B completed at 133 on both sides, with the unrelated lock held. Only head starts B's follow-up (events 137/140) before that lock is released. Both screenshot pairs are byte-identical 500 ms apart.

Full recordings: [Before](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/abe5fe9346b967a6/before-full.webm), [After](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f916770dc660839a/after-full.webm).