Linear rewards mean something like 10 SP vote = $0.01 and 100 SP vote = $ 0.10 which is 10x SP for 10x rewards (just giving example of relative scale, not absolute number). Under modest superlinear rewards, it means something like 10 SP vote = $0.001 and 100 SP vote = $0.03 which is 10x SP for 30x rewards and it'll exponentially increase. What this amounts to is the wide end of low SP users wouldn't be able to self-vote effectively and have a threshold before they get anything meaningful. It basically requires users to stack more votes for anything substantial, which promotes collective curation. However, the exponential increase would favour high SP users too much, hence the adding a cap to the modest superlinear, which means having linear rewards on the higher-end SP so that high SP users wouldn't get the disproportionate influence.
Yup, vote trading markets (and selfvoting) will become somewhat irrelevant under our proposal as it triggers a new economic equilibrium to equalise any style of voting (which is voting for others vs self-voting) and make it just as profitable no matter how one votes. Since there's also the modest superlinear rewards curve in place to encourage collective behaviour, more will tend to curate content. We can't eliminate self-voting, but we can make it less desirable to do so with new economic incentives that favour curation. It's like fixing a broken dam, but many can't see it at the moment because it's mathematical and has higher order effects on our activities here.
this can counter some of the illths of the inflationarity.