The fragments of large events split by the $changeStreamSplitLargeEvents stage might not be consecutive and might intertwine with other events from a sharded cluster. Something like this was already observed in the wild [TODO: reference needed]. This complicates the reconstruction of such events. To ease event reconstruction and avoid unnecessary complexity, we need to ensure that the fragments are prioritized by the merging algorithm and exit MongoS one after another without intertwining with other events.