Sad to hear this wasn't able to make it into GA.

But the link-relationship methodology is interesting (similar to something like PageRank through backlinks).

But it's not the methodology I'd have initially gravitated towards.

My first instinct would be relations based on subscription overlap. This seems like it should group commonalities based on the user interests. This may also have alleviated some of the SPAM issues.

Though it would have been interesting to see both approaches merged together.