When end users see disconnections, we are often at a loss to help them diagnose the problem because we don't have any client logs from the failed connection. We can ask them to run coder ping but if the issue is intermittent, the likelihood of a reproduction during the (short) ping is low. That's even assuming the issue is at the tailnet layer and not in the SSH code itself.

We should add an option to coder ssh such that the ssh client logs to a local file that the user could send us after a disconnect.