It's true, and just because we're not doing anything about it at the current moment, doesn't mean we shouldn't.
The way to go for votes and block messages is to see if a message of type m about round r was sent to us from node n. We should not be getting the same message type from a node for the same round more than once.
We can have an in-memory data structure for that.
The reason I commented, is that in replication, it's even easier to block it because if a node isn't actively replicating, it shouldn't be getting replication responses.
I think it's fairly simple to add it in the current PR, but we can also add it in a later PR if you feel this PR is too large.
Originally posted by @yacovm in #302 (comment)