There's a couple confusing things about this test/its history. Major highlights of the current behavior:
- On the classic jscore variant, this test succeeds without large windows of waiting
- Presumably changing this getmore into a readConcern: majority (as per what the suite does) with the same 1 hour timeout results in a 10 minute wait (cursor timeout?) before successfully returning.
It's unclear to me if the test is really passing. Certainly, it's not expected for the test to be waiting for 10 minutes before passing.
I accidentally let the test run to completion with debugging that exposes the contents of the returned batch. The one expected document exists: