If hashed_url mode is turned on and crossdb is also turned on, then queries to _memory should have a hash_id.

One way that it could work is to have the _memory hash be a hash of all the individual databases.

Otherwise, crossdb queries can get quit out of data if using aggressive caching.