If you add two separate Python nodes with identical properties, but no @id, to a Collection (e.g. if a given person is both an author and a custodian, but you create two separate objects rather than reusing one), then save the collection to a single file, there will be two JSON-LD objects in the file with different, local @ids.
We could consider having a deduplication step (automatic or optional) during saving, which will detect such duplicate nodes and merge them (i.e. keep only a single local @id.
If you add two separate Python nodes with identical properties, but no @id, to a Collection (e.g. if a given person is both an author and a custodian, but you create two separate objects rather than reusing one), then save the collection to a single file, there will be two JSON-LD objects in the file with different, local @ids.
We could consider having a deduplication step (automatic or optional) during saving, which will detect such duplicate nodes and merge them (i.e. keep only a single local @id.