The point being that I would expect stub article creation to be pretty high in new Wikipedias, and to tail off as the Wikipedia in question comes to be seen as more authoritative.
That's a very valid point. That would contribute to an upward trend of stub ratios with increasing edits, which is exactly what I see for Wikipedias with less than 20,000 edits, but after that it levels out, and becomes much less variable. I can't think why.