Never used it but afaik you could vote along different axis. So you could vote "+1 insightful" for instance, so if a comment had a lot of votes, you had a more idea why it was upvoted.
There was also meta-moderation, which allowed the community to deal with bad upvote/downvote behavior - brigading, stalking, etc. It wasn't perfect, but it did avoid the near-universal ills of moderator bias (including unconscious/cognitive bias), burnout from over-exposure, and flat out anti-empirical theories about how to do moderation properly. I wish more experiments like that had been tried (kuro5hin also made some efforts and Advogato took it even further) instead of basically reverting to a Usenet-era model.
I think the really valuable part was that you only got a few upvotes to hand out, and only once in a while, so you'd treat them with a lot more reverence