Stuck on http://cros-full-0010.mtv.corp.google.com/afe/#tab_id=view_job&object_id=230380956

The job got SIGABRTed

job_aborter is supposed to bring the job back to a good state however.  The logs indicate that it did so, and poking it so that job_aborter does it again marks the job failed as expected.

I'm not sure why job_aborter doesn't work sometimes; I can only assume it's somehow racing with the scheduler or shard client.  It's become quite clear that the scheduler does not properly create transactions when it's fiddling with the database.

I don't have a viable action item for the underlying bug, but fixed this host.