Topics that come up on phys-l often seem to inspire otherwise
reasonable people to take black or white positions and to argue as if
one can only do the physics by thinking about it one way--their way.
I've been guilty myself.
I would argue that this is one of those topics. While I happen to be
fundamentally in the "magnetic forces do NOT do work" camp, I'd be
lying to suggest that I don't routinely work problems as if they do.
Within the realm of classical physics and in a world that does not
include magnetic monopoles magnetic forces are ALWAYS and ONLY
exerted at right angles to the velocity of moving, electrically
charged objects. The fact that they, therefore, "do no work" is a
trivial and direct result.
Nevertheless, take the standard example of a current running through
a rod supported by frictionless horizontal rails in the presence of a
uniform vertical magnetic field. I venture to say that few if any of
us would balk at calculating a magnetic force on the rod, finding
that it is in (or against) the direction of motion of the rod, and
determining that the rod gains kinetic energy equal to the work done
by that magnetic force.
One CAN argue (correctly in my view) that the force that accelerates
the rod is an electric force that arises because of the Hall
polarization of the rod. (The negative conduction electrons are
driven preferentially toward one side of the rod and they pull the
positively charged lattice toward them.) In my view it's very useful
to see that one can look at it this way and reconcile the fact that
"magnetic forces do NOT do work," but it isn't a very convenient way
to solve problems.