Inconsistent treatment of pods that are scheduled but pending

[kdelee]

When submitting kubernetes work, at times a pod is created but not yet scheduled. This means the pod will not have all the containers started immediately. Most of the time receptor deals with this fine, showing the work status as Pending, which is the desired and correct behavior.

Occasionally I have seen the work unit end up in an Error state and have the error detail "Error creating pod: expected 1 container in pod but there were 0"
this means we are ending up in

receptor/pkg/workceptor/kubernetes.go

Line 222 in bfea897

return fmt.Errorf("expected 1 container in pod but there were %d", len(kw.pod.Status.ContainerStatuses))

I've found this when working on making awx respect more the state of the kubernetes work state for container group jobs in ansible/awx#11551