Rather than port the Bulk API into the CRUD Spec, let's just remove the bulk api from the existing bulk spec. That way, in the same way the find/getMore spec deals with implementation details, let the bulk spec deal with the implementation details of bulk writes and the CRUD API deal with a public API.
—
Original:

The CRUD spec relies on the deprecated Bulk API spec to describe a number of its features and semantics, particularly how operation-merging, ordered vs. unordered execution, result reporting, and error reporting work. Also, talking to old and new MongoDB servers and mixed clusters. Since the Bulk API is deprecated, it'd be nice if the relevant bits were pull from it and merged into CRUD, or perhaps published separately as a Driver Write Operations spec or something.