> If we do that then we still have a problem with overrunning memory
> after a sufficiently large number of tuples. However, that'd improve
> the constant factor by at least an order of magnitude, so it might be
> worth doing as an intermediate step. Still have to figure out whether
> the triggered-data-change business is significant or not.
I think that was part of the misunderstanding of the spec. I think the
spec means it to be within one statement (and its associated immediate
actions) rather than rest of transaction. I think it's mostly to
prevent loop cases A row 1 modifies B row 1 modifies A row 1 modifies ...
However, I only looked at it briefly a while back.