After thinking of it, I think it's unlikely the issue of "just rewards" to be solved by a single-dimensional algorithm.
Meaning, that it will probably require a multi-tiered approach by combining things like
a) author long-term "reputation"
b) author ratings by large or appointed curators who may do this manually
c) post ratings with something like a star system which build (a)
d) voting
...then all these should be weight-combined so that the first page displays not the highest earner (unless asked so by the drop down filter), but the higher combined quality/voting.
Whether segmentation of interest groups plays a role, I don't know. What I do know, is that the algorithm will have to evolve - and it will.
Another idea that sprung to mind, was the "request" of a user to be "evaluated" for quality. After writing his post down and before pressing submit, user pays a fixed amount of STEEM (like 1-2-5 steem) to apply for a "curators review" within the next 30 minutes or so... (money goes to curator or curators - no matter what the curator decides), and the post then gets a "quality multiplier" depending what the curator decided. Then the quality post starts at a different "base" level compared to the normal posts and thus attracts more attention from the get go. In case the curator-(p)reviewer misbehaves and the quality of the post is crap, a flagging/downvoting will remove a multiple of what the curator gained.
The thing about adding complexity is that if done sloppily, it can make it very much easier to game.
I've been thinking that a curation service something like what you describe could arise endogenously as it is. A whale could basically offer to look at peoples' posts for a specified fee, and then upvote the best of them. It would be crude, but effective.