My (dated) knowledge of AOA sensors is that they're more-or-less windvanes on their sides, so the air passes over them and pulls them straight back. If they're at an angle then the aircraft is not travelling directly forwards through the air, but is tail-sitting. Or nose-sitting, but between gravity and engines that won't last long before it's straight again.
So you have a mechanical component that rotates, and a sensor to detect the rotation. I expect they'll use Hall sensors so there's no contact and no bits to get mucky *inside*, but if the rotating component freezes, or gets caked in grease or muck so it can't move, it'll read wrongly.
If it seizes during landing (or on rotation at take-off), it'll be raised, and will register a high AOA. Nose-down to correct...?
If you had two operating at once, you could at least tell that you were getting inconsistent readings. If you had three, you'd have a pretty good guess which sensor to disregard.