This code in cluster is problematic:

            topology.update_max_election_id(description)
            topology.update_max_set_version(description)

... since it mutates the previous topology. Figure out where to put the logic in those methods so that it can be invoked without mutating the old topology. Maybe just return the new values.