Comment 4 for bug 1251920

So, it looks like the console log isn't growing...

Things I've checked:

 - I've downloaded the logs for the last 25 failures, and there's no commonality about what clusters fail, its spread
 - the cirros image we're using hasn't changed recently

Next idea, I'm going to bring up a devstack instance and see if I can get instances to fail without tempest.