Mongosync relies on being able to move data across two major releases (e.g. 4.4 -> 6.0). We currently test the update language across one major version.

We should add fuzz testing which applies generated updates to verison X and then re-applies them with $_internalApplyOplogUpdate on version X+1 and X+2, ensuring that the two clusters have identical results.

If any semantic differences are found, we should report them to the mongosync team, and determine whether it's possible there's an inconsistency in the field.

We should also add comments to the v:2 update application code that the semantics must not change in order to ensure data consistency.