That all sounds quite likely but i'm afraid i haven't been into the articles code recently as i always find there's something wrong with them and building a custom solution in trackers is generally far more flexible.
However, can i suggest a couple of things that will make it easier to test your theory.
- Reproduce it in trunk (will be Tiki 19 so all fixes should start there)
- Add a wish/bug report to the dev wishlist (and maybe set up a show instance, but debugging notification bugs is often tricky)
- Attach an svn patch or diff file of your suggested solution on the bug report
Sorry to ask you to do so much, but i just don't have time to experiment on this (hence the delay in replying, apologies)