I wonder is it possible to add coalescing optimization to a sequence of $group and $limit in aggregation pipeline?
For example, I have a pipeline like this:

db.messages.aggregate([
    {$match: {"company_id" : ObjectId("4c2118ad54397f271b000000")}},
    {$sort: {"ct": -1}},
    {$group: {_id: "$ti", ts: {$first: "$ct"}}},
    {$limit: 10}
])

My goal is to get small subset of $ti ("thread id") that have latest $ct ("conversation timestamp"). Execution time for this request is prohibitively high and I guess it's because $group stage process all input documents and apply limit afterwards.

It seems that incorporating $limit into $group processing would benefit this query by avoiding computations that will be thrown away in any case.