Ths context is that testing is a solved problem. You are taking specific factual examples I wrote and attempting to debunk them with a general statement that is in agreement with what I wrote. Testing scrollbards can absolutely be tested inside a VM. Pretending that native hardware is needed to cover 99% of the use-cases for scrollbars is kind of silly.

> Even my craptastic company gives me a bunch of devices to test on. And I very deliberately specify very low-end versions of the machines so that I can test against worst-case-scenarios. It's a bonus that sub-optimal hardware is cheap as chips.

You've made my point. This is a solved problem.