I am exposing codex as a simple cmd pass-through to WSL so that it runs in WSL as recommended by OpenAI. The app does not seem to give WSL enough time to spin up and invoke codex before deciding that it has timed out. There also doesn't seem to be any way in the UI to configure an allowable spool-up time to account for this.