Except that thread somehow couldn't get the same SQL to work in PostgreSQL. While here, Domingo claims this is working fine in PostgreSQL, and SQLite does strive for compatibility in that area. Thus that matters. But indeed, as long as divergence and limitations are clearly documented, it's completely OK for Richard to reduce the surface area of this new feature, which is also a welcome addition from my end. If Richard could actually explain how it works, and more precisely how that *magic trigger* allows gaining `RETURNING` support, I'd be glad. I know he's a super busy person, but I really wish for more deep dives from him on SQLite subjects, like SQLite internals. He is such an interesting and amazing mind.