During testing https://issues.redhat.com/browse/ETCD-328, identified a scenario where 

1. etcd cluster members is down to `1` member
2. there is no guarantee that this member will be `leader` 

The testing scenario involved scaling down by only one master-machine/ etcd-member at a time. However, when a replacement is being created by CPMS to replace the scaled-down master machine, and influenced by failure domains in https://github.com/openshift/enhancements/blob/master/enhancements/machine-api/control-plane-machine-set.md, the etcd cluster could end up with only one member, and no guarantee that this member will always be a leader.

Such scenario could result in corrupted etcd cluster state that does not reflect the current system state. IMO this scenario should be prohibited.

See https://github.com/openshift/origin/pull/27496#issuecomment-1293749080 for logs