Regarding trolls in general, and especially troll feeding, a couple of hackers figured out how to deal with them ages ago (in German): Throw comments into a bayes filter (like those spam filters), have them rank it according to what it learned and, most importantly: Don’t just block the comment. Have the user solve a captcha. The more incited the thread is, the more the post looks like trolling or responding to a troll, the more often randomly fail the captcha.
Thus you make the unwanted behaviour annoying, people often simply say “nah it’s not worth it”, but on the flip side you’re also circumventing accusations of censorship. Trolls, now having lost the massive reaction they so crave, and themselves having to jump through captcha hoops, migrate to darker pastures.