Not to mention the importance of stupendipity – the lucky discoveries you make thanks to your woeful ignorance of common knowledge in the field, or just sheer bone-headedness.
It leads me to believe that STEM fields shouldn’t be so selective in weeding out below average students. So if you increase the population of researchers/tinkerers, the probability of one of them stumbling across something incredibly useful rises. Obviously it depends on the field though and the cost of individual failures needs to be minimized.