If we could scan multiple underlying extents in parallel, we could parallelize some of the aggregation operations in aggregation pipelines. The code that would combine the results of smaller parallel tasks that results could be used in both mongos and within mongod when parallel cursor scans are available.