If you correct for the fact that the best researchers have zero incentive to teach (i.e. by just selecting those researchers and lecturers who are internally motivated to teach regardless of incentives) the researchers are way better on average.
There's a place for great researchers who suck at teaching or are so great at research that teaching is a waste, e.g. the institute for advanced study. (And frankly, most of the researchers at the institute should probably be teaching the best grad students.) But, as someone who was a grad student after going to a fancy undergrad school, I think it's clear that having most researchers teach is a fantastic Good Thing.
Which is perhaps the greatest bane of capitalism - that which has poor returns in the immediate time frame is always passed over for instant results. It hurts our long run dearly to be so inefficient.
How about that the person using the tool every day in the course of their work is probably going to be the person best placed to teach others how to use it? Obviously teaching is a skill and craft that has to be taught, but given two people of equal teaching ability, the person using the tool every day should be the better choice, surely?
There is no reason whatsoever why the same person should both create new knowledge and also teach calculus/cs 101/etc. I was good at research, but I sucked at teaching [1]. In contrast, many people are really terrible researchers, but actually pretty good teachers. (As you note, there is a glut of such people, so their wages are low.)
Separating these roles is a good thing.
[1] Due to a lack of standardized tests at the college level, I have no way to actually know this. But I do know my students didn't like me, complaining that I was too hard and unsympathetic.