He caved to the majority, directly in contradiction with better overall judgment (which is the role of the designer) and did so at the very end of development when such radical changes to something as essential as core game stats should be anathema (poor management.) If you told Tim to swap out the entire game engine less than 3 months before shipping, he'd tell you to pound sand. As a programmer, he knows that's not only stupid, it's suicidal. Why was his conviction on the design side lacking by comparison? Why did he go with the knee-jerk "solution" anyway? Why not propose (or implement by fiat, because the buck stops there) a compromise solution? Change the sliding scale of increasing point costs, make it cheaper overall, but preserve the exponential increase. Going from exponential to linear isn't a small thing. It would be far easier to iterate this as well, since the established pattern should have been set by previous testing at max "difficulty," giving an upper bound.