We introduced streaming $group for time-series collections when $group _id is on time under SERVER-70267. We could take this further for the firstN accumulator by:

1) Short-cutting accumulation for a single _id value when N results are found

2) Extending this optimization when _id is null, short-cutting when N results are found

As part of this ticket we should also consider the following, and either include in scope or create new tickets for them if they make sense:

1) Allow this short cut mechanism for firstN outside of time-series when possible

2) Extend the time-series use case to lastN when/if possible (this would require a reverse scan)