Codex app crashes on stale discovered SSH host after ~/.ssh/config is emptied or host entry is removed #18974

What version of the Codex App are you using (From “About Codex” dialog)?

26.417.41555 (1858)

What subscription do you have?

Pro (5x)

What platform is your computer?

Darwin 25.4.0 arm64 arm

What issue are you seeing?

Codex App crashes when a previously discovered SSH host is removed from ~/.ssh/config.

I was not using a Codex-managed connection that I added manually in the app. Instead, Codex App automatically discovered an SSH host from my ~/.ssh/config.

After that host had been discovered by the app, I removed its entry from ~/.ssh/config. Codex SSH discovery then correctly reported zero aliases, but the app still tried to render the previously discovered host: remote-ssh-discovered:lookme-server.

Instead of gracefully removing or invalidating that stale discovered host, the app hit an error boundary and showed a generic error screen.

Relevant log excerpts:

• [remote-connections/discovery] discovery_completed collectedAliasCount=0 collectedAliases=[]
• Failed to load Windows sandbox mode error="Error: No AppServerRequestClient registered for hostId: remote-ssh-discovered:lookme-server"
• Failed to load config errorMessage="No AppServerManager registered for hostId: remote-ssh-discovered:lookme-server"

Workaround: restoring a valid SSH host entry in ~/.ssh/config made the app behave normally again.

What steps can reproduce the bug?

1. Add a valid SSH host entry to ~/.ssh/config.

2. Launch Codex App and let it discover that SSH host automatically.

3. Open or select that discovered SSH host in Codex App.

4. Remove that host entry from ~/.ssh/config.

5. Refresh or restart Codex App.

6. Observe that SSH discovery reports zero aliases, but the app still tries to use the previously discovered host.

7. The app then fails with errors like:

   • No AppServerRequestClient registered for hostId: remote-ssh-discovered:lookme-server
   • No AppServerManager registered for hostId: remote-ssh-discovered:lookme-server

8. The UI falls into an error state instead of gracefully invalidating the stale discovered host.

What is the expected behavior?

If a previously discovered SSH host that was auto-discovered from ~/.ssh/config is later removed from ~/.ssh/config, Codex App should gracefully reconcile that stale discovered host out of the UI/app state.

Expected behavior would be one of:

• remove the stale discovered host from the remote connections list
• mark it as unavailable or invalid
• show a recoverable error message telling the user that the SSH host no longer exists in ~/.ssh/config

The app should not try to render a stale discovered host that no longer has a registered AppServerManager / AppServerRequestClient, and it should not fall into an error boundary or show a generic crash screen.

Additional information

This appears to be a stale discovered-host state reconciliation issue rather than a general SSH connectivity failure.

Log:

2026-04-22T09:51:47.979Z info [remote-connections/discovery] discovery_started entrypointPath=/Users/my_user_name/.ssh/config excludedAliasCount=0
2026-04-22T09:51:47.981Z info [remote-connections/discovery] discovery_completed collectedAliasCount=0 collectedAliases=[] dedupedConnectionCount=0 droppedAliasCount=0 entrypointPath=/Users/my_user_name/.ssh/config excludedAliasCount=0 filteredAliases=[] resolvedConnectionCount=0

2026-04-22T09:52:20.102Z error [electron-message-handler] Failed to load Windows sandbox mode error="Error: No AppServerRequestClient registered for hostId: remote-ssh-discovered:lookme-server"
2026-04-22T09:52:20.104Z error [electron-message-handler] Failed to load config errorMessage="No AppServerManager registered for hostId: remote-ssh-discovered:lookme-server" errorName=Error
2026-04-22T09:52:20.200Z error [electron-message-handler] error boundary ... errorMessage="No AppServerManager registered for hostId: remote-ssh-discovered:lookme-server"