CVMFS access failures
Since the migration to Kubernetes the cvmfs access has been really spotty: it seems like a significant fraction of jobs fail to connect, but then work after a few retries. Not sure if we are setting something up wrong or we should report this somewhere. @svanstro any ideas?