As part of the query_shape_hash_stability multiversion suites, we create two clusters with different versions using the MultiShardedClusterFixture. In this suite, we introduced an override which tries to get all nodes from both clusters through the DiscoverToplogy utilities. However, it seems that when ReplicaSetMonitor would try to parse the uri from the second cluster, it returned the mongod's from the first cluster instead. This is because it maintains a cache where the replicaset name is the key. Since both replica sets used the default name in both clusters, it would map mongods from one cluster as  the topology for both.
As a fix, we can introduce the option to set custom names for replica sets, such that each cluster has a different name for their replica sets, without sharing the cache for the ReplicaSetMonitor.