In your article on Bayesian spam filters (I didn't read the essay since before joining YC aka several years) You mentioned rules based systems only work up to a point. Today it's Valleywag (I don't care about the site one way or the other) tomorrow your getting requests to ban half a dozen sites per day.
I think that instead of banning a specific site a better solution is to make the algorithm better.
I think you can't really rely on people to withhold their votes when the content is engineered to take advantage of the human psyche.
Could a bayesian filter really separate fact and fiction? Cross-correlated ancedote and rumour? Satire and sensationalism? Humans struggle. Maybe one day...
The community doesn't seem to be able to correct itself, so if the content starts to veer away from the site's original principles, some external force is needed to gently nudge it back on track.
All that said, I don't really have a considered opinion on whether or not blacklisting Valleywag is a good thing; it is not something that I care much about.