As the person who coded longbets.org, I find this line of argument irritating. If the rule is "never trust anybody who made one prediction that, with hindsight, turns out to be wrong" then you're basically saying, "only trust people too wily to give hard predictions." It encourages the pundit's disease of never saying anything substantive.
You're completely right, but if the question is phrased "should I trust this person on a futuristic prediction concerning matters of complicated technology" and the person has made some pretty silly errors in the past concerning matters of complicated technology then the argument holds a lot more weight.
If we were debating pure macro economics, even just pure economics, I might give Krugman's opinions a lot more weight but he's repeatedly proven to not only generate controversy for page views, but also to make off the cuff predictions about things he doesn't know about (while still throwing his full academic weight behind said predictions).
Somebody looked at that whole set of predictions: http://marginalrevolution.com/marginalrevolution/2010/12/pau...
It seems a reasonable record for a set of predictions that where "the point was to be fun and provocative", as the article you link explains.