The kind of training necessary for becoming exceptional is entirely possible, but not scaleable, thus it's not a practical solution to the problem. I think the example you give supports that - mentorship tends to evolve organically from companies you work at, the colleagues you work with, etc. It'd be difficult to imagine good mentorship being assigned, and succeeding, at-scale.

Even the processes for training competent engineers are only barely scaleable. Most competent engineers are still being trained through colleges. Even newly emerging forms of training don't scale that well - take coding bootcamps with limited class sizes and the low success rate of things like Codecademy.