This ticket originated from issue 9 from the "Known situations where contracts unmet" section of the MongoDB Server read and write semantics doc.

This issue applies to write operations using coll scan or _id index scan running in a transaction with "rc:local"/"rc:majority" or running outside a transaction.

Such write operations, when run concurrently with shard key value updates, can miss matching documents and (if the operation is a multi:true update) can write to the same documents multiple times, violating the semantics in the read/write semantics doc (linked above).

The goal of this ticket is to improve the implementation of such write operations so they don't miss documents or write to the same documents multiple times when there are concurrent shard key value updates running.