Messages are not written to codex-tui.log if Codex exits too quickly

What version of Codex is running?

codex-cli 0.21.0

Which model were you using?

oss

What platform is your computer?

Darwin 24.6.0 x86_64 i386

What steps can reproduce the bug?

1. Use a fresh environment (no preexisting configuration, ollama is not installed, etc.)
2. Install Codex from npm

``bash
npm i -g @openai/codex@0.21.0
``

1. Tail Codex's log file

``bash
tail -F ~/.codex/log/codex-tui.log
``

1. Attempt to run codex with the --oss flag and debug or trace logging enabled

``bash
RUST_LOG=trace codex --oss
``

What is the expected behavior?

A warning log entry should have been written to codex-tui.log.

What do you see instead?

Nothing is written to the log.

Additional information

It appears that probe_server function in codex-rs/ollama/src/client.rs#L79 should either log "Failed to connect to Ollama server..." or "Failed to probe server at...". Neither is written to the log file.

I'm not a Rust programmer, but while investigating this I found this guidance from the tracing-appender crate's documentation:

Note: _guard is a WorkerGuard which is returned by tracing_appender::non_blocking to ensure buffered logs are flushed to their output in the case of abrupt terminations of a process. See WorkerGuard module for more details.

It appears that the guard returned by the non_blocking call in codex/codex-rs/tui/src/lib.rs is assigned to a variable but never used. As a result, buffered log messages are lost when the worker is terminated.