Good point. I'm thinking of "dead labs", where a tenured professor runs out of grant money and can't secure any more, so students and lab staff gradually disappear on their own, since the only official long-term job is the PI's. One unproductive professor's salary isn't much compared to an entire office or building in the D.C. suburbs.

(Empirically, a professor doing X will continue doing X until retirement, regardless of the grant situation. But students can still opt out.)