Currently you have to do a rolling restart of the cluster to reset the services-alumni state. The ticket JIRA tracking this issue is AER-2757. I see that it is code complete, pending review at the moment.
When released, you should be able to issue the following command to clear this state:
We have experienced similar situation, our cluster (+10 nodes) has experienced/reported serious cluster visibility issues when one of the nodes became unavailable (offline).
We have replication set to 2 so you could expect that it should not be an issue.
Is these days any way to manually let Aerospike know about situation where one of the nodes went down and should be removed from the cluster?