I do not think this is a reason to not use replication. In this sense, no distributed system guarantees consistency. If a node goes down (particularly a leader!) while trying to distribute a new update, you are very likely to run into issues.

Replication != backup