For what it's worth, I run a/b tests in my work where I consistently see large caching effects. For example, when we compare an experimental version of something to the currently deployed version it's critical that we cache-bust the control group. And changes that improve caching generally make things significantly better.

Caching effects are much smaller on Safari with its per-domain cache, but even there if something broke caching for us it would be a high priority to fix.

(I work at Google on ad tags)