I may be wrong, but my understanding is it’s an ECM software limit if the car is not in gear. Without a load, torsion on internal parts is different than when under normal conditions and stuff starts flexing. If parts are stressed beyond tolerances, it wouldn’t be difficult to start smashing parts together that shouldn’t.
Think of it like holding down the eraser end of a pencil firmly on a table and giving the other end a jolt. The free end of the pencil vibrates wildly, sometimes even to the point where it can break free of the end that’s being held against the table. Now wrap the pencil with duct tape. The change in load with a similar force causes a change in the vibration frequency - slower oscillations, less potential damage.
A bunch of years ago I had the pleasure of being present at during UL certification tests at a company I worked for. One of the tests clamps a part to an random oscillation table. The amplitude of the vibration remained constant due to the mechanical limits of the table, but the frequency was gradually increased until hitting the natural resonating frequency of the part. It was both shocking and spectacular to see the parts literally tear themselves apart at a particular frequency. It was almost if they had exploded. It was so cool to see I would have loved to become a UL certification engineer to break stuff like that for a living.