I think that in this case you get a non-trivial performance gain with some basic implementation improvements. You can replace the row-by-row round trip with a single SQL UPDATE with a LIMIT (or a LIMITed subquery if the server does it that way).

Also, the performance of the algorithm in the essay would be linear to the number of rows requiring update, can you elaborate on whether the RETE alogorithm can do any better in this case.

And finally, thanks for the note on RETE, I will have to investigate that.