Thanks for the reply. Your insistence convinced me to build the little geometrical model and check its implications.
Two kinds of motion are possible. Translation and rotation, and the center of rotation's location isn't clear.
With translation, the analysis is simple: at magnification = m, translating the camera/lens assembly distance d moves the image at the film plane by d. Translating the subject distance d moves the image at the film plate by md. Either way, focal length doesn't enter.
With rotation (assume for simplicity that the center of rotation is the lens' rear node), rotating the camera/lens assembly by a small angle theta moves the image at the film plane by (rear node to film plane distance)*tan(theta/2). Rear node to film plane distance given focal length f and magnification m is f*(1 + m) so rotating the camera/lens assembly by theta moves the image at the film plane by f*(1 + m)*tan(theta/2). In this case focal length does matter.
Whether focal length matters in practice depends on how camera shake is divided between translation and rotation. In my experience, shake is all translation, no rotation.
Note, however, that I shoot closeup handheld with a Nikon and flash illumination. This approach eliminates motion blur and, until KM went away, allowed full control over aperture. Measurement trumps theory.
What is your experience? If you shoot closeup at all, how do you light your shots?