I’m reluctant to add new (replacement) etcd nodes until I understand why

kubectl get no

is misreporting.