Here’s the last experiment I’ve made (had the idea just today):
I wrapped the sensor in the lightweight plastic bag, and put loads of paper towels inside to block the sensor’s line of sight. It looks like this:
The idea is to block off any real sources of motion for the sensor to detect – and it does, if I walk around and under the sensor, there are no detections any more – and any further motion detected by the sensor would then be a sensor hallucination.
And indeed, it still hallucinates motion.
Honestly I should’ve tried this from the beginning. It seems like such a simple test and it immediately removes a plethora of different possibilities as plausible causes of the problem…