The point about raising the voltage and lowering the current is an approximation that you get when you assume that the motor is 100% efficient...another thing that doesn't actually happen in the real world
Because motors are often in the 90%+ efficiency range, it is actually a pretty good approximation around the nominal operating point.
Since real motors are not 100% efficient, then the voltage versus current characteristics will not follow an ideal constant power curve.
For any given drive frequency and torque, the motor will have an ideal operating voltage at which it is most efficient. So on top of the real world 'non-constant load' breaking our approximation, you have real world 'non-constant efficiency' breaking the approximation.
A real world induction motor at its nominal operating voltage and _no load_ probably draws somewhere between 30% and 60% of its nominal full load current. This is almost all reactive current (no real power delivered), and is drawn to create the motor magnetic field.
The relationship between the magnetizing current flow and the applied voltage is very non-linear, and strongly depends upon the saturation of the iron. With 480V applied to a 240V motor, I would expect so much magnetizing current to flow that I am very surprised that no OCPD or overload popped.
Is it possible that the motor was in fact wired for 480V, but with a coil reversed, or with the rotation reversed?
-Jon