While users can specify batchSize as part of cursor options when opening $changeStream as well as when issuing getMore request, the batchSize attribute is respected only on the router, but not on the shards. The batchSize is not specified as part of the request to the shards. Customers have requested this for having more control on how much work is being done on the shards with the purpose of decreasing the event latency time.

Slack thread: https://mongodb.slack.com/archives/G01FJN0RYVA/p1741938065747779

The original issue describes situation in $changeStream, but the issue is most likely not limited to it