Recovery of a Couchbase cluster when one or all nodes fail

Hi There ,
I have gone into similar issue which is described in below link, and was able to resolve the issue after lot trobleshooting.
randomly before reading below issue i was able to resolve the issue in similar steps.
but my question is now what might be actual reason for this kind of issue ?

https://blog.jtclark.ca/2014/07/simple-recovery-of-a-couchbase-cluster-when-one-or-all-nodes-fail/

In my case i had 3 node cluster and restarted all the 3 nodes in rolling fashion , all nodes were good and rebalanced after rstart.
but all of a sudden tow nodes were in down , one node is pending other was down . for these two nodes, they are exchanging the status as down <=> pending rebalance. but i dont have an option to rebalance.
after a final decision i thought to failover one node then the other node was back to cluster with active status.
and then i brought back the failed node into cluster.
But i couldn’t figure out the reason.
version : Couchbase 4.5