Pool rewards are a function of r_shares.
They used to be non linear (x^n), then in HF19, they were changed to linear/proportional.

Simplified Example:

Right now my vote is worth 0.90$ - with no regards to how many other people voted the content. It is 0.90$ when I am the only one voting for a post, it is 0.90$ when the post is already sitting at an expected payout of 100$.

Before, it was hard to predict the value of a single vote, because it also mattered, if other votes were active on the post. So just a rough example: 0.20$ when I am the only one voting, but on a 100$ post my vote could be worth 2$.
This would also mean, however, that my downvote on a 100$ post would deduct 2$.
More eyes, more impact.

The problem was, that the whales are lazy and don't like to read, so they all voted on the same post via autovote, which lead to just a few authors getting all the rewards.
But that was a problem with collusion and general ignorance, not the code itself.

I do believe however that only a few posts per day should get rewards, because the average user will only read a handful of posts per day in the real world.
The top content gets by far the most views and should get by far the highest rewards.

bla bla bla I could go on, but this seems too complicated for most people anyways and I am annoyed right now, sorry.