The optimized path of create collection consist in creating and pre-splitting the chunks across the cluster depending on the split policy initially associated with the collection.

SERVER-59806 introduced an optimization when creating a collection in the optimized path: using an insert instead of an upsert of the chunks. However, we could also set the ordered field to false to allow parallel batch processing instead of sequential processing and decrease the total time necessary to pre-split large amounts of chunks in the cluster.