will.buerger@mongodb.com suggested adding a generic $sort optimization (via a rule that lives on the $sort stage) that erases itself (the $sort stage) when it sorts by the latest sort pattern. This ticket should investigate the feasibility of this and based on that determination, implement it/add the appropriate unit/integration tests.

More details excerpted from Will here:
"In other words, if stages are responsible for reporting metadata about document sort-key population, then we can register an optimization with the $sort stage that introspects the sort-key populated once that stage is reached (a la dependency analysis). Related to my comment about making this optimization apply to $search as well, this would make this optimization apply more generically

That also might help with sharding? Since we're looking behind instead of ahead, we don't need to worry about the shard-side losing pipeline suffix dependencies. $sort ends up in the merging pipeline so it only requires analysis pre-pipeline-split"