Deceptive ray tracing bug

Brandon Denning in my ray tracing class showed me an image for his specular reflection assignment, and I told him it was definitely right:

But it was not!  Turned out he was using the direction of normal vector to send the "reflected" ray.  It's interesting how plausible it is (to me anyway).  Here's the right one:

There is a related bug in Kajiya's original path tracing paper that fooled him and the rest of us for years.   The worst bugs are the plausible ones!