CursorEstablisher has some logic to decide which failure to return in the event that it sends multiple requests that fail with distinct errors. It appears that it is possible for it to prefer an error that is received when a sleep associated with a retry backoff is cancelled, possibly as a result of the original error.

As an example, this can be encountered with the following sequence:

• CursorEstablisher sends two requests
• First request is rejected by the shard's rate limiter, inducing a backoff sleep
• Second request fails with a CollectionUUIDMismatch error
• First request is cancelled via stopRetrying, yielding a CallbackCanceled error ("Baton wait canceled")
• Final error returned is CallbackCanceled, instead of CollectionUUIDMismatch

We should update CursorEstablisher's error handling logic to ensure that the original error is not lost in these situations.

https://github.com/mongodb/mongo/blob/ab099727112ab36c8972c47d12145ce4e3462b11/src/mongo/s/query/exec/establish_cursors.cpp#L380