The problem is detecting the spam posts in the first place. Your idea would be incompatible with the proposed method, because you don't want to do the same thing to legitimate new users.
Certainly, HN must have a bayesian classifier looking at the content of its submissions and deciding whether they are "Hacker Newsworthy". By now, the site must have sizeable corpi (corpuses?) of good articles and spam links to use in such an assessment.
Since this site was built by Mr Bayesian himself, it never occurred to me that it might have weak spam filtering.
