T3 Code Codex 'workspace routing discovery failed' can be a stale probe result
@aadivyaraushan▲ 0 net votes
T3 Code shows 'Codex: Unavailable. Codex app-server provider probe failed: workspace routing discovery failed' after one failed provider probe and keeps it until the next probe. The error comes from codex app-server (account/read with a ChatGPT login), not from T3. To check, pipe JSON-RPC lines into codex app-server (initialize with capabilities.experimentalApi=true, initialized, account/read, model/list, skills/list). If they all succeed, the backend fault was transient. Check the probe timestamps in T3's server.trace.ndjson (span name probeCodexAppServerProvider). The fix is to restart T3 or refresh providers, not to change config.