External Publication
Visit Post

Codex extension crashes with SIGKILL; last log shows slow INSERT into logs

OpenAI Developer Community May 16, 2026
Source

Hello,

I’m seeing the Codex IDE extension crash with the following error dialog:

(code=null, signal=SIGKILL)

The most recent visible error says:

“slow statement: execution time exceeded alert threshold”

The dialog shows what appears to be an internal Codex log write:

INSERT INTO logs ( ts, ts_nanos, level, target, feedback_log_body, thread_id, process_uuid, module_path, file, line, estimated_bytes ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)

The elapsed time shown is around 1.105s.

This does not look like it is coming from my project code. It appears Codex is being killed while writing to its own internal logs database or cache.

Environment:

  • OS: macOS
  • IDE: [VS Code / Cursor / other]
  • Codex extension version: [insert version]
  • Codex CLI version: [output of codex --version]
  • MCP servers configured: [yes/no]
  • User config present: ~/.codex/config.toml [yes/no]
  • Project config present: .codex/config.toml [yes/no]

What happened:

  1. I opened Codex in the IDE.
  2. Codex crashed and showed the error dialog.
  3. The dialog suggested checking config.toml, updating IDE settings, or downloading a different version of the extension.
  4. Reloading the extension [does / does not] temporarily fix it.

What I have tried:

  • Reloaded the Codex extension: [result]
  • Checked for updates: [result]
  • Temporarily moved ~/.codex/config.toml: [result]
  • Temporarily moved project .codex/config.toml: [result]
  • Ran codex --version: [result]
  • Tried opening a different project: [result]

Question:

Is this a known issue with the Codex IDE extension, local log storage, or MCP/config handling? Is there a recommended way to clear the local Codex cache/log database or collect more diagnostic logs for this crash?

I’ve attached a screenshot of the error dialog.

Discussion in the ATmosphere

Loading comments...