Started on 12/2, this isn't 100% failure because it's a race
to make calls to c-vol before it dies on each startup, but it
is the root cause in a recent spike in volume-backed server
delete timeouts in the multinode grenade job.
Change-Id: Id5936ae52305fa1906bb7a5d6bed863c2d922a0a
Related-Bug: #1647789