I'm curious what might be pathological cases, patterns of query watching and updates that give the cache a lot of work to do to keep up