I noticed this in the stainless results here:

https://stainless.corp.google.com/browse/chromeos-autotest-results/251352302-chromeos-test/

The test for the webserver on localhost:8000 failed and in the log I see this:

2018-10-23T16:29:30.229522-07:00 INFO vm_cicerone[26403]: Startup of container penguin at IP 100.115.92.199 for VM termina completed.
2018-10-23T16:29:31.359805-07:00 INFO vm_concierge[26386]: Received GetContainerSshKeys request
2018-10-23T16:29:31.438866-07:00 INFO vm_cicerone[26403]:  Process 10 exited with status 255

The 'exited with status 255' from vm_cicerone can only be caused by the SSH process terminating...but it launched correctly because otherwise there would be a different error.

This has only happened twice in the past month of testing, so it's quite rare. Once on relm and once on eve.

Here's the other occurrence:
https://stainless.corp.google.com/browse/chromeos-autotest-results/245988086-chromeos-test/