Based on SERVER-84347 which was supposed to make DISTINCT_SCAN optimization available to $top and $bottom accumulators, it seems to rely on this form of group:

{$group: {
    _id: "$a", 
    accum: {$top: {
          output: "$b", 
          sortBy: {a: 1, b: 1}}}
}}

But no one would know to write that - took me a few minutes to figure it out:

{$group: {
    _id: "$a", 
    accum: {$top: {
          output: "$b", 
          sortBy: {b: 1}}}
}}

is the normal and expected syntax that we should expect to optimize.  Of course it would rely on index on a and b which is its own little side issue ...