The Hot Network Questions sidebar is a bit of a contentious issue, because it tends to promote questions that are often perceived as 'cheap', of relatively low quality, or on the fringes of topicality for the site that hosts them.
One reason for this is that they're subject to a feedback loop: they're promoted widely, a bunch of people with no previous experience of the site click through, upvote it (because they bring their association bonus) but don't downvote it even if they recognize it as bad (because the association bonus doesn't let you downvote), and then the question score increases, and the system thinks it's really Hot and worthy of further promotion.
This has frustrated a lot of people, to the extent that people have asked that outsiders with just the association bonus be barred from voting on those questions (or, for that matter, asking how to get rid of the sidebar entirely). That, however, is a bit extreme, and makes understandable that the dev and community teams have given that proposal a wide berth over the two years since it was first proposed.
So, here's something a bit less extreme:
Let association-bonus users vote on HNQs, but don't count those votes and views into whatever algorithm does the decision that a question is Hot.
That is, separate all the votes and views on questions and answers into two categories,
- those by users with 100+ rep earned on the site in question, and
- those by everyone else, even if they have 100 rep from the association bonus,
and discard the second category when feeding it into the HNQ algorithm (possibly still this one?). (If you want to, you can change that 100 to some given
threshold, no smaller than the earned ability to upvote at 15 rep.)
I think this is a good mix of not being too disruptive on the normal mechanisms while also allowing more room for the organic quality measures of the individual sites to decide what gets promoted more widely, instead of the current junk-food driven algorithm.
That said, I would at least like to know from the dev team whether there are insurmountable scaling and feasibility barriers to implementing this sort of separation.