Later votes are just as valuable for rankings and for essentially confirming the judgment of early voters.
Yes, but unless there is a significant penalty for later voting, there is a huge incentive to pile on.
For example, if curation rewards were 25%, and they were just divided up evenly according to stake weight without regard to who voted first, late voters would have a huge incentive to pile on to already successful posts for their share of a 25% of guaranteed moon.
This is a secondary consequence of extreme superlinearity. Taking the extreme opposite case of fully-linear rewards, even without any further adjustment, there would be no incentive to pile on. The curation rewards for being the last voter would be the same as voting on any new post.
I'm pretty sure that even if superlinearity in some form is kept we can in principle invert the superlinear function such the the last voter does not get a huge bonus from the superlinearity, but also doesn't get nothing.
Thats the problem with talking about changing curation rewards before we know exactly exactly where we're going w/ author rewards.. but yeah the piling on thing is gone if there's no superlinearity.
Of course, if there's no superlinearity and no significant reward for voting early, then it doesn't really matter that much what you vote on at all.
Significant is relative. I don't believe that the degree of order-penalty in the current system that essentially gives no reward at all to later voters is necessary. Earlier voters can still get a bonus. For example, perhaps 10% of each voter's reward could be paid to earlier voters as a finders fee. The first voter (getting fees from all of the rest) would earn by far the most (assuming several times as much total voting as that one vote), but even the last voter would earn 90% of the base reward.
In general I agree that the basic shape of rewards would need to be understood first, most likely.