Damn good article. I knew that table variables didn't have stats computed for them, but I hadn't fully considered the implications of that.

One point which your article doesn't make explicitly is that table variables live outside transactions. I realize that you show this in the comparison table at the bottom, but I think an article as good as this one should give this point at least a mention in the Other Differences section, because this behavior is 1. potentially useful and 2. an (admittedly) rare source of difficult-to-find bugs.