I do understand your point of view, but perhaps I'm just not optimistic on the ability to figure an ultimate ROI formula that incentivizes good behavior. I feel like cultural change and UI improvement is more important going forward.

Maybe I'm just missing how some of these suggested changes even work to properly address the issue at hand. In my mind, most folks who earn via vote selling are doing so by delegating power to a service that then pays them a percentage of their earning in turn. These voting bots are then dominating curation rewards as well, and under an increase in curation % will adjust by passing on more of that profit back to their delegators, thus keeping vote selling in equilibrium with actual curation.
And it doesn't seemingly do anything to address multi account abuses. Any kind of collusive voting ring will see no impact to their activity as it'll just shift which account involved gets the ill gotten gains. Likewise pure self voters remain unaffected as it'll still be more profitable for them to vote themselves as gaining the curation reward and the creation reward is better than just curating.

Maybe my thinking is off on some of these points, so please feel free to correct or hit me with some examples, but overall, I believe that whatever math we do, there'll be follow up math to determine the best way to game the system. Even with changes to make rewards non linear again it'll still be a solvable problem for ROI maximalists resulting in regular users still not getting curated. Voting trails would simply return in force to pile on predetermined posts at the most mathematically beneficial time to maximize rewards. For a small segment of the population, proof of brain will always end at the point the numbers are crunched and the bot is programmed.

All that being said, I don't think it's hopeless to expect changes and better things for Steem. But that kind of gets me back to my thought that some of these discussions may take us off course. For example, I firmly believe the best way to undercut the voting bots is to replace and improve the existing inbuilt promoted & advertising system. A well implemented and working system would give people a choice, and go a long way toward undercutting the profitability of vote selling services and improving content discovery, as well as bringing users a secondary revenue stream that would actually be valued based on their genuine audience size & engagement, rather than the amount of Steem Power they currently hold.

Perhaps I favor a more bottom up approach, rather than top down. I'm kind of working off the assumption that the "power users" are going to find ultimate ROI at the expense of the spirit of proof of brain no matter what. On the flip side, the vast majority of average users don't even fully grasp the mechanics of the rewards system as is, but they're the mass consumer feeding the vote buying industry because they see no alternative. They're the ones whose behavior can most be changed to improve the platform. Peer to peer advertising programs, public education about the importance of powering up and exercising your votes, and perhaps on the technical side continuing to look at the problem of the "dust threshold" taking away voting power of the smaller users. I just find things like this as being more important and productive than trying to calculate a new rewards formula that will maybe improve the voting patterns of a couple dozen individuals.