Filtering on topic name tends to help a lot in very big systems (I've seen OpenSplice systems with 100k readers but only up to small 100s or for a specific topic). Since the point of this PR is that it fixes observed performance issues, clearly this is only something to keep in mind rather than something that needs to be changed in this PR.