> the other needs more design work than a small amendment.
I agree that one seems more difficult, I'll also think about it but so
far I don't have a better idea than what you suggested.
> I'd rather propose a design for this separately, likely a session-scoped
> set of "types altered in this transaction" plus
> a check at composite-read time. Happy to open a fresh thread on it.
That's fine with me if others agree with the approach, the current
patch is a clear improvement over the current behavior.
Case 2 is the only remaining issue I can reproduce currently, I can
crash the server with multiple variations with it, but those all
follow the same generic idea.