I balanced a prism using the single-plane 2 run method. 
Let's look at a plot of the polygon motor with AN44000A chip and an unbalanced 30x30x2 mm square prism. 

A clear peak is present at 50 Hertz for both the acceleration and infrared spectrum.
The rotor is pulsed at a frequency of 10 Hertz. It apparently spins five times faster.

Before adding a weight the following measurements are obtained;
Adding a weight at the exact opposite location reduces the accelerometer amplitude.
The rotor tends to jump to different frequencies. The NBC31111 is more stable.  For a frequency of 30 Hz the NBC31111 needs to be pulse at 180 Hz. The AN44000A needs to be pulsed at 5 hertz to achieve 33 Hz.

Mirror motors have 5 pins; PWM, MOTOR_EN, GND, Voltage and LCK_PIN.
PWM is speed control. GND is ground. Voltage is the supply voltage, ideally 24V. LCK_pin is
false if encoder and signal are not in sync. I cannot confirm the function of LCK_pin and MOTOR_EN.
Changing them or looking at their input does not reveal any activity.
I will try to expose with the AN44000A. So far I only exposed with NBC31111.

Finally, Max who bought the first prisms found some interesting links.

Baraja
Baraja is building a LIDAR scanner which uses a prism with an uneven number of facets.
They do not rotate the prism but change the frequency of the laser.
https://www.baraja.com/

DIY linear motor
There is an instructable of someone who build a DIY linear motor. In the first machine with a prism scanner a linear motor was used.
https://www.instructables.com/DIY-IRONLESS-LINEAR-SERVO-MOTOR/