ET doesn't notice when SSH_AUTH_SOCK changes underneath

[Yostage]

On our systems the SSH_AUTH_SOCK can expire and get changed by the system underneath. When this happens, starting other SSH connections from your et session will just hang forever. If you know what's happening, you can work around by opening a new tmux window.

I wonder if there's a more durable fix by invalidating the cached auth sock or managing this turnover in some other way inside et.

[tekknolagi]

This happens to me when I do the following pattern:

• Use et from Machine A to log into my server
• Open tmux and set up some long running thing
• (I'm not sure if Machine A has to drop the connection or if it doesn't matter here)
• Use et from Machine B to log into my server
• Reconnect to tmux. See the long running thing failed. Try sshing somewhere from inside tmux
• See that the connection hangs Even though I am not using et from within tmux, the stale SSH_AUTH_SOCK causes a hang within ssh. It's probably better to cause an error since then ssh will move on to another method.

cc @tchebb

EDIT: We also have a reasonable suspicion that this behavior started happening with the security updates in Nov 2021, but have not confirmed.