Like a lot of people here I started seriously looking at buying thermal clip-ons in the last couple of years, and at the start, I heard a lot about how the best systems were collimated with a Risley prism. My current understanding is that that is largely accidental misinformation carried over from the world of night vision (image intensification) clip-ons. I will explain my current understanding below, but I am happy to be corrected by an expert if one comes along.
A Risley prism is basically a clever arrangement of wedge-shaped prism lenses that allow you to "steer" a light beam or path. See here for a neat demo with a laser beam:
YouTube link. Although the laser beam comes out of the emitter in a straight line, you can see that you can alter its path very precisely by rotating the prisms to a certain position. Some high-end NV clip-ons like the PVS-30 use such a system. It is not user-adustable, but during the manufacturing process, each device is attached to a test fixture and aimed at a calibrated light source, and then the prisms are set such that the image seen at the ocular end is the same as what goes into the objective end. Then the setting is locked down, and the result is that the image is basically passing straight through the intensification clip-on without being distorted or bent. Thus, as long as the day optic and NV clip-on are reasonably aligned, there will be minimal POI shift when the clip-on device is attached.
Thermals are different. The light is not really "passing through" the device in the way that light passes through an NV tube. The infrared waves fall onto a microbolometer sensor (or a fancy pixel-counting sensor for cooled MWIR systems) and are stopped completely there. For the clip-on to function correctly, you really just have to ensure that the displayed image on the rear screen matches what the day optic would have seen when the clip-on is removed, but this does not require any "bending" of the light path at all, so no Risley prism is necessary. Instead, special software calibration systems are used to ensure that a given pixel on the microbolometer activates the correct pixel on the display screen on the back of the unit when it is hit by infrared light. They also have to ensure that the image
size is calibrated correctly, so that 1 mil on the day optic reticle is 1 mil when looking through the clip-on, and not 1.002 mil. This is factory set for military clip-ons primarily to ensure the device will work with minimal POI shift on just about any weapon if the day scope and clip-on are reasonably aligned. However, there is no real reason that a civilian end-user couldn't fine tune the screen image position for their exact rifle, and indeed, many thermal clip-ons allow this. However, doing so ruins the ability to slap the thing onto any rifle in your squad and have it work well enough to engage the enemy without re-calibration. Since the military doesn't want to futz with calibration, the extra cost of pre-calibrated units makes sense, but for civilians, that extra labor isn't necessarily worth the extra price tag, so it can be omitted on civilian units. The owner is then forced to calibrate the clip-on for a given rifle or a few rifles in his collection. Many clip-ons intended for military use do not allow the owner to even access a calibration menu, to prevent grunts from fucking up the calibration.