PR 22425 real environment transcript
Date: 2026-10-30 UTC
Main: 32ecd71726fee16f335484c1f95cbce126d461f6
PR head: 67d41c66adf2ffce5de9d775e9c4328dce560d50

Main: git worktree add ../main github/main; vp i
Started background dev server, captured PID 30983. Ports 16733 / 8693.
Created project with t3 project add.
t3 thread start --base-dir .t3 --provider codex --model gpt-6.1-sol ...
Result: completed; MAIN_PROVIDER_OK
Stopped main before PR server.

PR: checked out exact head, vp i, background vp run dev --home-dir /home/user/t3code/.t3
Ports 13773 / 5733. Final server captured PID 39597.
Created real git repo /home/user/usage-hold-repo and project through project add.
Chrome /usr/bin/google-chrome; Playwright 1.60.0 from repo dependencies.
First real turn: completed; PR_PROVIDER_OK
Send active turn: timeout 100 sleep 90, then respond.
Send second message through t3 thread send: delivery=queued, status=queued.
Steer active turn through web textbox + Enter: provider returned CANCEL_ME_UI.
Queue started automatically after active success: provider returned QUEUED_TURN_OK.
Final thread status: completed; runCount=3.

Limit-specific reproduction: NOT VERIFIED. No exhausted account available; real Codex turns succeeded. No synthetic usage-limit events, fixtures, mocks, or database writes were used to simulate quota.
Known-reset expiry, unknown-reset hold, and switching away from an actually limited instance were not exercised.
Linux web/server only; native desktop/mobile, remote/tunnel not exercised.

Second provider instance: added codex_2 through Settings > Providers > Add provider > Configure manually. Same local proxy credentials, not an independently exhausted account.
Selected Codex 2 / GPT-6.1-Sol through web model picker and submitted SECOND_INSTANCE_OK prompt. Persisted run ordinal 4 completed on codex_2 at 11:27:25.668Z; prior three runs remain completed on codex.
Checks reported by cafe on exact head: 29 successful, 0 failing, 0 pending. No source changes; no new test files, commits, or pushes.