Eyeballing this, the effect size is tiny. Looking at their own measurements, it is statistically significant, but barely.
ADDED: Hmm... I missed the second page. Over there is more explanation of the analysis. In particular:
But this analysis gives us a way to quantify the advantage to Baskerville. It’s small, but it’s about a 1% to 2% difference — 1.5% to be exact, which may seem small but to me is rather large... Many online marketers would kill for a 2% advantage either in more clicks or more clicks leading to sales.
Point taken. This is large enough that it might be useful. However, I don't think it is a large enough bias to be important for rationalist.
My A/B test has finished: http://www.gwern.net/a-b-testing#fonts
Baskerville wasn't the top font in the end, but the differences between the fonts were all trivial even with an ungodly large sample size of n=142,983 (split over 4 fonts). I dunno if the NYT result is valid, but if there's any effect, I'm not seeing it in terms of how long people spend reading my website's pages.
If it's worth saying, but not worth its own post (even in Discussion), then it goes here.