Recently, I discovered a shard that was not offloading
job and task results because of bug 684122.  There were
no alerts to say that there was a problem.

We need metrics that can measure the forward progress of
gs_offloader, and alert if there's a problem.