Surprising to see the extended discussion surrounding the effort to fit
a road mirage into the ray optics paradigm. Some of it mentioned total
internal reflection in this connection.
But wait! Suppose I examine a situation with two adjacent
parallel rays approaching a region of reducing refractive index at a
slight angle. Call them A & B. The first ray curves gently away, and
above it, the parallel ray closely follows the first ray's trajectory.
Approaching they are ordered A & B, departing they are also A & B.
Now suppose I consider two such rays approaching a mirror surface:
first the lower ray reflects up, then the second ray likewise. In
approaching the rays are A and B, on departing B and A. An inversion
of a familiar kind.