Richard F. Wintle describes his job coordinating grant-seeking and laboratory work in a Canadian research institute: "The unsung heroes behind those big genomics breakthroughs". The subheadline is "Sometimes the best way to do an experiment is to have someone else do it for you".
Perhaps I'm fortunate in that I'm not an independent, "Principal Investigator" researcher, meaning that I largely don't have to teach, and I usually don't write my own grants. A big piece of what I spend my days doing is helping other scientists get their applications in shape, and even more importantly, helping to run a core facility to help them get their experiments done. It's a slightly unusual niche for a career scientist – not an independent researcher, not a lecturer, but more like running a small-to-medium sized biotechnology company that happens to be not-for-profit.
Science is done by those who don't teach much, and that has consequences for the kind of science we do, the kinds of questions we ask.
I was a participant in a panel here yesterday helping early-career faculty become better teachers. It is really sobering to realize just how little teaching most faculty members in the sciences do. From my context in anthropology, it is hard to imagine a new Ph.D. going straight into a tenure-track faculty position with no prior teaching experience whatsoever. Fields in the biological sciences are really varied in what teaching obligations they expect from faculty, and from my students' standpoint, it shows.
I think there's no question that teaching makes people better researchers, but there's also little question that teaching obligations make faculty publish less. There's a trade-off. I think we should prefer better research as opposed to more publications. But the linked article gives a different perspective -- the ultimate step being the "outsourcing" of research from universities altogether.