Due to historical reason (e.g. inserts via direct connection) or unforeseen range deleter bugs, it may happen for documents to exist on a shard not owning the range they fall into.

In order to prevent mixing up legit documents (incoming via a migration) and spurious ones (incorrectly present within the range being moved), we could add a check at the beginning of cloning and fail the migration if documents are already present. This would be as cheap as opening a cursor on the shard key index at the beginning of the range and checking the shard key of the first returned document.