Voting cartels can not be mitigated or minimized completely, even after so many forks they will find some way to survive. And sometimes we may not be able to establish abuse objectively. That's one fundamental problem.
Second, I personally believe that self upvote feature should be disabled in the frontend, and at the "0" layer adequate adjustment should be made.
Third problem is that an established blogger after some period, like 2 or 3 years or on account of his good behavior historically used to get some momentum votes just after posting(trail votes/automated votes), so the votes in such a case do not read the content, but on the premise of hitherto established reputation, they consider it as an acceptable content without reading it.
There exists a wrong conception of Hive Power since it is directly obtained by powering of HIVE, users do have a preconceived notion of Power to maximize APR, they do not take it as an influential power in the first place. Influence power comes with a responsibility to sustain that influence. If abuse of reward pool is going on, then unknowingly the people are weakening the worth of their own HP and influence power.
Decentralized curation (proof brain) may need a hybrid mechanism with some adjustment, it may also be a complex one. But as it evolves it has to fine-tune many such things simply for the reason that we intend to make it more human than AI, and human behavior is quite complex.
As engagement is the cornerstone of Hive to improve the odds of curation/vote to their posts, some people might game it with low or spam comments. And may further web a complex cliche, thye may try to maximize it to feature in Hive engagement league report. So checking the nature of abuse in the comment section if a herculean task, not because they would directly spam, but because they might tweak the things bit to masquerade it as genuine engagement. And cringe comments could be another headahce.