Comment 3 for bug 1850189

I saw this occur again on SM-4. I think the next steps here is to come up with a reliable reproduction scenario. I think this can be achieved with a test to run consecutive lock/unlock cycles. What I observed, after an unlock, was that the k8s API was not accessible from logs in sysinv within the same timeframe as the tiller error. We might need to determine if there is intermittent API access on the AIO-SX since there is only one replica. If intermittent access is not an issue, we may need to test the API access prior to running the application-apply as after a host unlock the apply might be happening before all the pods are are recovered.