All payouts that earned very little already are reduced even more with this curve.

No theyre not. Vshares aren't porportional to payout except as a percentage of total vshares assigned.

as an example, imagine two posts. One post has 1MV of SP voting for it. The other, 100MV of SP lets assume that 1MV = 3 rshares, and and 100MV = 300 rshares.

Under the current system, the first post would get 3^2=9 vshares and the second post would get 300^2=90000 Vshares.. As a result, the second post would be paid 10000x more than the first post.

Now lets look at the modified system. Its dependent on the threshold, but lets take the n^2/n-1 idea that steemit, inc came up with (because i think its a better formula)

Under that system, yes, youre correct, the absolute number of vshares on the 3rshares post would be less. it would go from 9 to 5.5. But the second post, the one with 300 rshares would go down even more (from 90,000 to 909). As a result, the first post would get just under 1% of the reward pool, and the second post would get just over 99% (which is pretty close to linear, where it would be exactly 1% and 99%).

Note that, like others, i think this threshold is far too high. (though, tbh i like the one proposed by steemit inc)

It reduces payouts across the board. All payouts that earned very little already are reduced even more with this curve. It doesn't make the distinction between grandma's travel post or someone who would abuse the system.

Its not really possible to "reduce payout accross the board" (without changing the composition of the reward pool) because, regardless of the number of Vshares assigned, there is still the same amount of money being distributed.

The number of Vshares a post gets doesn't really matter. Its the percentage it gets of all vshares.