[rocqua]

Honestly, when mastery of the subject becomes detrimental to teaching the subject, your not taking the highest-level courses. A great researcher has to teach and convince themselves, those skills translate to teaching at that level.

I'm mostly thinking about the highest level math courses though.

[vanderZwan]

I guess you're talking about the point where it's more collaboration on a half-solved problem than data transfer about solidified knowledge? Because then I agree, but that's a really different type of knowledge transfer we're talking about then.

[rocqua]

There's a point before collaboration, where you ask a researcher to bring a student up to the level of the researcher. At that point, the student and researcher struggle with similar enough problems that the researcher can think from the students perspective.