It's fun to hear what worked and what didn't for complex web rendering. It would be interesting to see some graphs or numbers behind the performance gotchas listed in the article.