...
B records with a 'null' A-key value would be silently dropped
Implications for many-to-many relationships
to resolve many-to-many relationships using the infrastructure build for one-to-many relationships and additional group by and map step is required. The Output of the first one-to-many join had to be grouped by B's key. The aggregate would keep track of all joined records that exists and have different A keys. A Map phase would select a record at random. The proposed one-to-many join can then be used to join with the additional entity.
Compatibility, Deprecation, and Migration Plan
...