PR #30674 real runtime verification

Baseline github/main: 8b44f42b14c8e273156474a924ccf29680dd7857

Original PR head: de32a70542b0917cf3c7ee920d99754ea1d38d42

Corrected head: 00de53dacccd4583e7a4e3d9920640870dd0d879

Client: real Vite dev web app, Playwright dependency, /usr/bin/google-chrome.

Provider: Codex gpt-6.1-sol via local proxy. Data: app-created project and thread, real hosted PR comment 6626636678; no host responses mocked.

Running-server reads: GET /api/orchestration/shell and GET /api/orchestration/threads/<threadId>, using the authenticated browser session and protocol header 2.

On main, new comment generated a real notification/provider turn; subsequent edit produced no edit notification. The GitCafe host returned updatedAt but its adapter discarded it.

Original PR after edited comment and a normal sweep:

{
  "observedAt": "2026-11-04T08:55:40.022Z",
  "provider": {
    "instanceId": "codex",
    "model": "gpt-6.1-sol"
  },
  "run": "run:thread:31ec9ac1-a75e-4ce1-87fe-01f7037f7247:ordinal:2",
  "remarksThrough": "2026-11-04T08:54:05.381Z",
  "pendingEdits": [],
  "wakes": 0
}

Corrected server first held edit:

{
  "observedAt": "2026-11-04T08:57:30.416Z",
  "provider": {
    "instanceId": "codex",
    "model": "gpt-6.1-sol"
  },
  "run": "run:thread:31ec9ac1-a75e-4ce1-87fe-01f7037f7247:ordinal:2",
  "remarksThrough": "2026-11-04T08:54:34.750Z",
  "pendingEdits": [
    {
      "id": "prcomment_01m6jvx3pzedw9btp41z11x7gk",
      "editedAt": "2026-11-04T08:54:34.750Z"
    }
  ],
  "wakes": 0
}

Next normal sweep, newer edit held with no provider turn:

{
  "observedAt": "2026-11-04T08:59:18.748Z",
  "provider": {
    "instanceId": "codex",
    "model": "gpt-6.1-sol"
  },
  "run": "run:thread:31ec9ac1-a75e-4ce1-87fe-01f7037f7247:ordinal:2",
  "remarksThrough": "2026-11-04T08:57:32.181Z",
  "pendingEdits": [
    {
      "id": "prcomment_01m6jvx3pzedw9btp41z11x7gk",
      "editedAt": "2026-11-04T08:57:32.181Z"
    }
  ],
  "wakes": 0
}

Following normal sweep, second newer edit held with no provider turn:

{
  "observedAt": "2026-11-04T09:01:17.360Z",
  "provider": {
    "instanceId": "codex",
    "model": "gpt-6.1-sol"
  },
  "run": "run:thread:31ec9ac1-a75e-4ce1-87fe-01f7037f7247:ordinal:2",
  "remarksThrough": "2026-11-04T08:59:27.615Z",
  "pendingEdits": [
    {
      "id": "prcomment_01m6jvx3pzedw9btp41z11x7gk",
      "editedAt": "2026-11-04T08:59:27.615Z"
    }
  ],
  "wakes": 0
}

Settled edit and following normal sweep:

{
  "at": "2026-11-04T09:05:20.830Z",
  "watch": {
    "startedAt": "2026-11-04T08:54:05.381Z",
    "headSha": "de32a70542b0917cf3c7ee920d99754ea1d38d42",
    "failedChecks": [],
    "passed": true,
    "passedChecks": [
      "Check",
      "CodeRabbit",
      "Mobile Native Static Analysis",
      "Rust",
      "Test Web",
      "Build and Typecheck",
      "Release Smoke",
      "Test",
      "Sync PR size label definitions",
      "Label PR size",
      "Label PR 30674",
      "Deploy relay preview",
      "Deploy web preview",
      "EAS Preview",
      "Native fingerprint diff",
      "Test Server 5",
      "Test Server 1",
      "Test Server 8",
      "Changes",
      "Test Server 4",
      "Test Server 2",
      "Test Server 7",
      "Test Server 3",
      "Test Server 6",
      "Lint",
      "Agent context budget",
      "Collect PR targets",
      "Prepare PR size config"
    ],
    "owedChecks": [],
    "remarksThrough": "2026-11-04T08:59:27.615Z",
    "remarkIds": [
      "prcomment_01m6jvx3pzedw9btp41z11x7gk"
    ],
    "conflicting": false,
    "wakes": 1
  },
  "providerRunCount": 3,
  "messages": [
    {
      "at": "2026-11-04T09:03:22.789Z",
      "text": "acknowledged: 1 edited comment from maria-rcks. model: gpt-6.1-sol."
    },
    {
      "at": "2026-11-04T08:55:09.026Z",
      "text": "Update on pull request #30674 (https://git.cafe/pingdotgg/t3code/pulls/30674), which T3 Code is watching for you:\n- All 28 checks passed on de32a70.\n\nLook into each item and act on it as your task requires. This PR stays linked and watched, so do not link, watch, or list it again. End your turn when you are done, and T3 Code wakes you on the next change. When you hand the work back to the user, call unwatch_pull_request first so the thread returns to their inbox.",
      "summary": "#30674: checks passed"
    },
    {
      "at": "2026-11-04T09:03:16.409Z",
      "text": "Update on pull request #30674 (https://git.cafe/pingdotgg/t3code/pulls/30674), which T3 Code is watching for you:\n- 1 edited comment:\n  - maria-rcks: \"> [!NOTE] > Written by `gpt-6.1-sol` on behalf of Maria ## Verified in a real dev environment Runtime verification: 165 scoped tests pass. The first changed edit was held without waking Codex; now ...\"\n\nLook into each item and act on it as your task requires. This PR stays linked and watched, so do not link, watch, or list it again. End your turn when you are done, and T3 Code wakes you on the next change. When you hand the work back to the user, call unwatch_pull_request first so the thread returns to their inbox.",
      "summary": "#30674: edited comments"
    }
  ]
}

UI Stop action:

{
  "at": "2026-11-04T09:05:58.937Z",
  "watch": null,
  "pendingBackgroundTasks": []
}

Blacksmith red/green: existing adapter test failed with expected undefined to equal 2026-09-12T12:02:00Z; all 25 adapter tests passed after the fix.

Blacksmith vp run check:changed HEAD~2 (original PR + correction): format, lint, both knip passes, GraphQL, typecheck (36 affected packages), 165 tests across 5 files passed.

Unverified: direct GitHub watch (only available commenting account is excluded there), native desktop/mobile, live paginated review-thread tails, host read failures and 10-wake limit (covered by existing automated tests).