And here's a theoretical underpinning to explain so it doesn't come across as simply a subjective design choice: coloring objects darker or lighter with different shades of the same base color to show intensity is one of the few principles I remember from an Edward Tufte workshop from a few years ago. Varying levels of saturation indicate a shift in degree, whereas different colors altogether imply that the objects are of differing qualities.
Applied to the new commenting feature, darker colors would imply more "solid" feelings of trustworthiness, similar to how trolls already get washed out with the lighter shades.
Similarly, that explains another thing wrong with the highlighting - "alarm" colors like orange, red, yellow etc indicate that something's wrong, whereas bolding or darkening the names wouldn't imply that at all.
Trying to suss out good and bad comments reminds me of the 37signals Troll Cap and the "Good Comment" crown:
http://www.37signals.com/svn/archives2/introducing_the_troll...
http://www.37signals.com/svn/archives2/the_crown_of_royalty....
Not too important but, to quote our house dev motto, "behavior contrary to specification mandates a bug report no matter how minor the deviation is".
[Edit: I refresh this page and am seeing it sometimes highlighted and sometimes not highlighted, so I'm not sure if my bug report is accurate or not. Apparently the Heisenberg Uncertainty Principle applies to bugs: attempting to describe one changes its behavior.]
i bet discussion interestingness improves by removing the username and points.
the advantage of seeing points is the same as seeing orange usernames. it allows people to decide which comments they want to read.
I think if you're going to read, you might as well read everything that isn't grayed out with downvotes.