If a fast cpu masks a "problem", I'd submit that it might not be a problem. Hardware is cheap, I try to save my brain cycles for stuff that matters, like HN, .emacs.d, and codegolf!
The "hardware is cheap" phrase was rampant in our company until about a year ago. A year ago an executive took a closer look at our cloud bill and, well, exploded. Now there's a huge push to reclaim those extra cycles and RAM as our "cheap hardware" bill is outrunning our overall salary costs.
Being at a point where this is a problem could be considered to be a good thing; we have enough traffic to cause it to be an issue. The problem is, the MVP was cost efficient (even at scale), it's the re-write that's costing us money. At some point, you have to grow up and realize that throwing hardware at poor code is not a sustainable solution.
At scale, the equation flips. Hardware becomes expensive, and any given developer's time is cheap.
I believe this is a great choice. Most of our code will live in the cloud, and will be billed by the second. If you can feel if something goes slow, you can fix it before you're billed for it ;-)
Fast hardware can mask performance problems. I still remember a bug I had in a Turbo Pascal code when I switched from my 286 to a Pentium: my loop counter had an extra '0' and it run unnoticed on the pentium for hours because it was too fast...