I'm using a dummy collection of 18M rows in the following format:

{ _id: __unique_of_course__, name: __also_unique__, dummyID: __not_unique__ }

There's an additional index on dummyID:
db.dummy.ensureIndex(

{dummyID: 1}

);

Now, when I query it ...
db.dummy.count(

{dummyID: __not_unique__}

);

This will take several seconds ....

Now, when I query the following ...
db.dummy.find(

{dummyID: __not_unique__}

).sort(

{dummyID: -1}

).limit(10);

This takes just one ms or so ...

Is this some bug, or did I just forgot something?

It'ld make sense if sorting a 18M result-set took more time than counting it ...

(note: it may be faster when testing it on a real server, I'm testing it on my MacBook right now. Look at the relative difference between the benchmarks).