To give context, till that point he'd explained that a uniform gravitational field can be simulated or is equivalent to going to another coordinate frame that's accelerating (constant accelaration in the opposite direction) irrespective of the position.
He then moves on to the case of a gravitational field pointing radially inward towards the sun:
There's a 2000 ft man who's falling legs first towards the sun. He'll feel a stretch since the force on legs is greater than that on his head. The claim is that no coordinate transformation can simulate this effect.
The way I'm trying to understand it is: if that person were in an elevator in vacuum, there'd be no kind of motion that the elevator can follow such that he'd feel a similar stretch across his body. My question is: why are we sure that there isn't some sort of super-weird motion that the elevator can follow, such that he feels the same stretching effect?
As an example, let's say the person was in the middle of a giant, hollowed-out cylinder (oriented along the cylinder axis). Then he'd feel a radially outward pulling force on his body. If we try to simulate this with the person in the vacuum elevator, no matter what translations we do (zero/non-zero, constant/non-constant accelaration), we can't replicate the radially outward pulling. But now I come to know of this very new kind of motion called rotation, and realize that if the elevator is rotated about its vertical axis, the person inside will feel the same as when he was in the cylinder.
So to reiterate (and I'm certain that I'm missing something) why do we say so surely that the radially inward example (where the person gets stretched) can't be replicated by a vacuum elevator?