Occasionally, build fails due to faulty hypervisors.

We’re looking into adding automated retry capabilities and eviction of misbehaving hypervisors to make sure these internal serverful failures are not seen from an end-user perspective.