It would be interesting to automatically detect when a thread is doing a scan that pulls pages into the cache, and pre-emptively make accessing those pages more efficient.

The goal of this ticket is to spend some time figuring out how hard that is and how beneficial it can be.