I think it is impossible to design the ideal rules upfront so experimentation is important to reiterate closer to something that will generate the desired behaviours. The voting I see is too focussed on earning through what looks like robovoting for authors that already have successful (high earning) publications. Good for them but on a system level that does not achieve the stated goal of finding good content. Established authors are already "found".

On the other hand, there needs to be some sort of predictability, if someone with hundreds of (new) followers is upvoted but that does not translate to earnings that will demotivate. It is not healthy that the only shot you have at a decent earning (10's not 1000's of SBD) is to be found by a whale. That simply won't scale and getting earnings with ever increasing content becomes more of a lottery at best and nepotism at worst...

Too much surprises or inconsistencies with the reward system and people will disengage.
I think it is better to experiment now and risk breaking stuff in the short run rather than allowing a faulty feedback system to persist which can cause much greater damage.