"I know that there's arguments that the two should stay entangled..."

I've just got one, but it's pretty compelling: the best instructors I've seen have been researchers teaching about the subject they're researching. That might be an introductory, undergraduate AI class taught by a senior AI researcher, or a new faculty member effectively teaching about his dissertation topic.

Sure, I've seen plenty of researchers who made bad instructors, but I've seen many more instructional faculty who not only didn't care about teaching but also didn't know the basics of their subject.

I'm not sure that it's not possible for top rate researches to teach at attached universities if they care to. There's plenty of professionals who work in the commercial world and teach a class or two a semester because they love it but hate working in the academic world -- I work with two PhDs who do exactly that.

Anecdotally, many of the worst teachers I had were researchers in their field. It was clear that they wanted to just research in their field and teaching was something the school forced them into doing and they did the bare minimum.

I've noticed that a few research professors have struggled to teach courses. It's not a problem of being unable to convey the information (since CS is conference-heavy in publications, you can't become a good researcher without getting good at presenting), but rather that they can't write do homework assignments or tests very well.

Imo at least the following important subjects would be too unprofitable for industrial research labs: philosophy, pure mathematics, particle physics, astrophysics, religious studies, philology, history, anthropology.

This might be useful for postdocs, but remember that the PhD degree is a research degree. Therefore, academia has to have some kind of research available for people to do. Assuming that the typical PhD in Chemistry anyway takes 4-6 years based on topic and level of support. The first two years are coursework and other prep work for entering the actual research process. Semester #5 is the prelim, which is the entry exam. The next 2-3 semesters are the golden time when other interests fall away, work progresses and the material that justifies the PhD gets done. After that comes writing up and doing whatever odds and ends the boss wants done or wants done to expand/round-out the dissertation or whatever.

It's possible that lab-based PhD's take longer because the reactions have to work and yield the desired product. Also, a lot will depend on how often the student's in the lab (reactions have to be worked up, machine time has to be found, etc). Run 2x vs 1x daily has a direct impact on time-in-process. Fortunately, the non-lab PhD's can work pretty much anywhere as modern laptops are either sufficient or work well as front-ends to compute resources.

The end result of this is 3-7 papers, which will be "stapled together" to make the dissertation. The work has to be "novel", and some people are lucky/good enough to produce something that's a win. Note that some of the best ideas in the field (possibly any field) come from 2-4th year grad students who don't know or care they're wrong... Oh, they might produce some code throughout this process, or some groups maintain a code base that pretty much every student contributes to (such as QM groups).

Coming out of this process, the newly-minted PhD is now capable of doing an academic postdoc to either round out their training or start the path towards an academic position. Or, they can do an industrial postdoc or go directly into pharma or biotech (or the various support companies doing chemistry, software or whatever supporting these two areas). Think of the PhD as a union card, certifying the holder as being (vaguely?) capable of independent, organized work. Lab PhD's typically post-doc with the "other half" of their discipline: come out as a synthetic PhD, go work for a natural product group, etc.

Once in pharma, they face a 3-6 year period where they're to learn various aspects of the medicinal chemistry process. They start as pairs of hands and end up, hopefully, as somebody who can run with at least part of a drug design or development project. Academia, for the most part, lacks the money or interest for this part of the training process - it's applied research with an eye towards the big return, and IP issues prevent publishing until projects die or go into the market.

I suspect physics PhD's might have similar timelines, possibly delayed if they have to build their apparatus (or wait for time at CERN). I've heard that UK PhD's get through quicker, and it's possible European PhD's get through a little slower than North American due to a better funding environment? I've no understand of the schedule for things like Biology or Medical-related PhD's as that's totally outside my experience (personal and professional). Might be slower as the mice/cells/whatever can up and die on them.

To close this spiel, while research labs have always provided a place for PhD's to go work (and still do), there's always problems due to security, funding, etc. If there was more money available, we'd end up with a better crop of postdocs and probably young professors with better home lives but that doesn't seem to be the trend of things right now. Given all the uncertainty over funding, finding a position, avoiding layoff/outsoucing/offshoring, ... I can see why people aren't interested in going this route.

Those that do either have a "cause" they wish to follow/fulfill or they're really interested in the subject matter and want to take it farther than one can do with courses. If people want the $$, they'll do far better managing or funding people like I've described than being one.