Intuition is probably not a good guide here. An improvement to the kernel build process is relevant to the thousands of machines that are used for kernel devel

Intuition is probably not a good guide here. An improvement to the kernel build process is relevant to the thousands of machines that are used for kernel development; an improvement to the page fault speed is relevant to the over 1 billion devices that run the kernel. That's a pretty big multiplier. (I believe the GPs reasoning is that "improve the build system" and "retard the build system" are not symmetrical, because both directions require positive effort to be expended).