When one side of a merge is treesame to the ancestor, we can take the other side and skip all the expensive merge operations. This optimization can only be performed when the generation of REUC extension data is skipped.
