In SPM-3509, we added logic to detect when a $search query will need all results from mongot due to a post-$search blocking stage ($sort/$group). When that is the case, our strategy for receiving all documents from mongot is exponentially increase the batchSize per batch, with a starting value of 101, and always prefetch the next batch from mongot.

We could consider starting with a higher initial batchSize value than 101 since we know we'll need all the documents. It may be worth running some benchmarks to check what starting value is most efficient.