The irony is that it often takes only a tiny number of test users to spot a glaring usability problem that could sink a major UI change. Someone (possibly the Nielsen Norman Group?) did some research on this a while ago and found that even single digits of test users would be enough to collect very useful feedback.

People are so obsessed with scale and techniques like A/B testing today that they seem to forget that kind of test can only tell you whether one of the options is significantly better. It can't tell you whether you're finding the best choice between excellent options or the least of evils between dreadful ones.