runtime verification, 2026-11-10, Linux Chrome + Playwright
main: 4c924ff4e6; head: a7c5392cc6851cda3edfb66a6a35fb07f6f0ab80
real local Whisper Tiny English; espeak audio played into a PulseAudio remapped source, captured by Chrome getUserMedia. no speech responses or provider responses intercepted. Codex gpt-6.1-sol via local proxy.

read-only persisted run query: /tmp/main-baseline.sqlite
{
  "thread": "63f4cdbb-82f3-4259-bb5a-cae1cd149129",
  "runs": [
    {
      "ordinal": 1,
      "model": {
        "instanceId": "codex",
        "model": "gpt-6.1-sol"
      },
      "status": "completed"
    },
    {
      "ordinal": 2,
      "model": {
        "instanceId": "codex",
        "model": "gpt-6.1-sol"
      },
      "status": "completed"
    }
  ]
}
{
  "thread": "6273ce9c-b576-4257-8409-9699da4bec52",
  "runs": [
    {
      "ordinal": 1,
      "model": {
        "instanceId": "codex",
        "model": "gpt-6.1-sol"
      },
      "status": "completed"
    }
  ]
}

read-only persisted run query: .t3/userdata/statev2.sqlite
{
  "thread": "63f4cdbb-82f3-4259-bb5a-cae1cd149129",
  "runs": [
    {
      "ordinal": 1,
      "model": {
        "instanceId": "codex",
        "model": "gpt-6.1-sol"
      },
      "status": "completed"
    },
    {
      "ordinal": 2,
      "model": {
        "instanceId": "codex",
        "model": "gpt-6.1-sol"
      },
      "status": "completed"
    },
    {
      "ordinal": 3,
      "model": {
        "instanceId": "codex",
        "model": "gpt-6.1-sol"
      },
      "status": "completed"
    },
    {
      "ordinal": 4,
      "model": {
        "instanceId": "codex",
        "model": "gpt-6.1-sol"
      },
      "status": "completed"
    }
  ]
}
{
  "thread": "6273ce9c-b576-4257-8409-9699da4bec52",
  "runs": [
    {
      "ordinal": 1,
      "model": {
        "instanceId": "codex",
        "model": "gpt-6.1-sol"
      },
      "status": "completed"
    },
    {
      "ordinal": 2,
      "model": {
        "instanceId": "codex",
        "model": "gpt-6.1-sol"
      },
      "status": "completed"
    }
  ]
}

main: confirmed Send + navigate B => no new run in A (2 completed runs, including control send that stayed in A).
PR: same flow => A run 3 completed once; B draft unchanged and survives reload.
PR: Settings round-trip without cancellation => A run 4 completed.
PR: Escape during pending send and Cancel voice input after Settings round-trip => no additional A run, still 4.
PR: normal typed send in B => B run 2 completed, reply normal typed send works.