Even if the astronaut were a rigid body, he or she could still rotate albeit inefficiently. For example, let the astronaut raised his hands over his (her) head, and consider this the initial orientation. Now if (s)he moved his arm forward, his legs would go backward. When the astronaut moved his arms over his head again, the astronaut would return to the initial orientation. Now if the astronaut moved his arms out to one side, the legs would go to the other side. When the astronaut returned his arms directly over his head, (s)he would again return to the initial orientation.
Now assume the astronaut moved his arms forward so his feet went backward. Now the astronaut moves both arms to one side, and his legs go the the other side. Now the astronaut moves his arms directly over his head again, (in one step; not the two steps (s)he took to get to the side). The astronaut would have his/her arms directly over his head, but (s)he would have rotated. This is because, as is well (should be) known to students of college freshman mechanics, finite rigid body rotations are non-commutative.
As a grad student, I briefly talked to some (other) grad students and they do not buy this argument, but at the same time they could not refute it.