Thanks for your reply!

Is it true that N^x does have anti-abusing mechanism where x > 1?

What about the idea of letting witnesses adjust whatever parameters are in whatever the reward function is?

For the existing simple n^2 equation, we could replace it with n^x where x = 1 + 1/y where y ranges from 1 to 10. The witnesses could simply vote for 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10. So the equation's flatness could be varied in 10 steps between the flatest n^1.1 to the current n^2.

Or if we go with modified n, any thoughts on letting witnesses vote on the exponent and constant in the denominator?