There is a relatively new statistic that tracks the amount of time internal threads spend waiting for the dhandle list lock. That statistic does not currently include time spent by the eviction server, since the eviction server uses a custom lock acquisition scheme for the dhandle lock.

We should enhance the statistics code to include tracking of time the eviction server is spending waiting for the lock.

This raises a general question about whether usage of trylock should be accounted for in the lock acquisition timing macros. WiredTiger has two different use cases for try lock, one if for cases that implement a custom backoff lock acquisition and the other is for operations that are only performed if a lock is available. I think it's interesting to track the former, but not the latter. Unfortunately by their nature the former aren't easy to group into a common function call, so I think we need to add tracking around trylock calls on a case-by-case basis.