Rob Hasselbaum That was my recollection too. However the original work does partially try to handle id values, although just for the read part. You can see it in your patch attachment as part of the changes to AbstractEntityPersister:

+	private final String[] rootTableKeyColumnReaders;
+	private final String[] rootTableKeyColumnReaderTemplates; 

I am just trying to determine why write fragments were not accounted for there. I could not remember if there was a specific discussion about those or what.

As for the actual HHH-9808 report, I can track that down. More than likely the readers/templates were just missed in a few places that SQL is rendered.