If the migration is the first sharding-aware operation that the recipient shard is participating in, then the _recvChunkStart command it receives will make it sharding-aware and load the chunk data for the chunk being migrated for the first time. If the collection was only recently created and sharded, there's a risk that it could load out of date data for the collection, causing it to see no chunks for this collection, which causes the migration to fail with an obscure "data transfer error". If the source shard included the config server optime when it sent _recvChunkStart, then there'd be no risk of the recipient shard reading stale data.