I agree with your assessment that the distribution mechanism is broken. When it comes down to changing user behavior or refactoring algorithms, it's clear to me that user behavior won't change. Nobody really flags. They complain that there isn't an incentive to doing so, but if you look at it long enough there is, but that if you can get beyond ROI-centric views. Most folks wont.

Which leaves us with refactoring the algorithm. In my opinion, we are suffering from a nothing-at-stake problem where users can behave in undesirable ways because they have little to lose. If we increase the downside (beyond flags) so that risk comes into play, then low effort milking becomes harder to do, because something is at stake. But of course such an idea would be massively unpopular because freeloaders gotta freeload.