Thanks, that’s really helpful. The “test the extremes manually” workflow is exactly what I was curious about.
I agree that subjective UX is difficult to automate. What I’m experimenting with is a bit different — rather than trying to decide whether the UX is “good,” it records a test flow once, replays the same flow at several resolutions, and looks for visual/layout regressions or unexpected differences for a human to review.
Would something like that actually save you time, or would you still prefer checking the extreme resolutions manually?