The full scenario is as follows:
We have two entities (Foo and Bar). Foo has a multi-value entity reference field to Bar entities.
The migration source outputs each Foo-Bar pair in a single CSV line (manual aggregation is not possible because there are thousands of entries).
My plan was to put several entries with identical keys and using the --update flag and a custom process plugin, update the entry's entity by adding the Bar entity to the existing Bar entities instead of replacing them.
When I tried to put two entries with identical keys in a single CSV file, the migration module processed the first entry and ignored the second.
My understanding of the migration module is that each entry is assigned a key and that that key is mapped to the entity that was created by the migration.
Is there a way force the migration module to process multiple entries with identical keys?