The goal is to use this as a cluster health metric. We expect these operations to be uniformly low-latency. The first step is to investigate what's already covered by serverStatus and queryStats, but my sense is that we are missing this observability.