Firestore in Datastore mode supports counting via aggregation query. Counting this way seems faster than the current methods used in NDB (skipping/brute force).

Can the current count() implementation use counting via aggregation query? (Note that this does not work on the datastore emulator)