I think the light bulb finally came on with the difference between driving a motor and driving an inductive or resistive load.
An inductive load will build up current and when that current is interrupted the voltage will reverse and become very large if the current has no path. In a motor this is not the major effect. The momentum of the rotor has a greater impact with the exact opposite effect. The back EMF from the motor coils turning in a magnetic field opposes the current flow. So if a switch is opened to stop the current flow initially the voltage will spike and current will flow in the same direction through whatever diode is available. As that current is dissipated the back EMF from the rotor movement dominates and creates a polarity of voltage that is the same as originally applied. This voltage is fixed for whatever current flows. If current does flow it saps power from the rotor motion and it slows.
So in a motor controller a PWM circuit will apply voltage and current to the motor in the appropriate polarity then open the circuit (with an inductive EMF quenching diode) allowing the motor to not be braked, but free wheeling.
In an inductive load the control would be a bit different. Using PWM to control the average current and power into an inductive load requires the "OFF" time to short the outputs to allow the current to continue to flow.
A resistive load would work with either of these approaches.
I'm looking at using a motor controller to act as a class D amplifier. If the load looks inductive I think a motor controller that opens the output FETs during the off part of the cycle would not work so well, dissipating power during that time rather than allowing it to cycle through a pair of FETs allowing the current to continue to flow.
So I guess the EMI filtering will need to be primarily resistive or capacitive.