As far as I see if I restart the failed compute a second time then it starts up properly. So a double restart can be a workaround.
As far as I see if I restart the failed compute a second time then it starts up properly. So a double restart can be a workaround.