I am open to other solutions ...
My idea is a linear curation rewards curve where it wouldn't matter when you vote and how many other users voted on (or are expected to upvote) a post (which would also remove the five minutes 'curation window'...).
I know there is the myth of the manual curator who spots quality content first and then benefits of all the upvotes to follow. However, in reality the big majority of early upvotes are automated whereas real manual curators get the smallest piece of the cake. I see not a single reason why I should get less curation rewards when manually upvoting a two days old post which I had really read and evaluated than someone who didn't work at all but just let a programm do the job!
I explained my idea more extensively here (where I had to defend it in the comment section). :)
To balance the disadvantage that self-voting (because of the higher, linear curation rewards) would be somewhat more profitable again, I think the author rewards curve could stay convergent linear, and in addition one moght think about going back to something like 75 % author rewards and 25 % curations rewrds.
However, with 'free' flags since EIP I am not sure if that would be necessary at all.