The initial retry up to 20s was giving up too soon.
I'm seeing this test flake because the Node rebooted and it takes ~2min
to recover.
Now StatefulSet RunHostCmd calls will use the same 5min timeout as with
other Pod state checks.