Let's say I have two documents in a collection:
{_id: 0, a: 1, b: 1} and {_id: 1, a: 1, b: 2}
and index on {a: 1}.
Aggregation pipeline [{$sort: {a: 1}}, {$group: {_id: '$a'}] will be covered by the index, however pipeline [{$sort: {a: 1}}, {$group: {_id: '$a', b: {$first: '$b'}}}], according to execution stats, will examine both 2 documents despite only needing 1 value for $first operator.
So this seems like a place for aggregation optimization.