I was bored at work today
...
The next step is to modify the 1/focal length rule to keep the effectiveness constant across all focal lengths. Any takers?
I have another suggestion for the case you are still bored and want to fine tune the recognitions from your thoughts.
Think of following experiment: you have a wall large enough to fill the image. The wall is plastered with some grainy stoff, the grains are of varying color; you go so far, that each grain is captured by one pixel. All lenses you considered are
rectilinear, thus one grain is one pixel in the center and in the farthest corner (apart from lens distortion, let's ignore that), assumed the lense's axis is perpendicular to the wall.
Now, the point:
one grain in the center occupies much more "angular space" than in the corner. Or, in general, the angle of view covered by one pixel is greater in the center than in the corners. Accordingly, the effect of camera shake depends not only on the lens, but on the location in the image:
the center gets much less blurred than the corners. Moreover, the rate of change from the center to the corner depends on the focal length: long lenses show a more uniform effect than wide angle.