In sharded search, the shards compute the min/max DocsNeededBounds of the shards' pipeline in order to tune the batchSizes requested from mongot. However, the min/max constraints used may not be reflective of the full user pipeline, since the shards don't have knowledge of the merging pipeline. Currently doing so would be very difficult due to dependency cycles (work tracked in SERVER-89546), but we should consider doing this at some point.