You write the software to detect and be resilient in the face of hardware failure.

If there is only one sensor, how do you know it’s reporting wrong data?

One sensor failed. If one sensor gives readings that don’t jibe with all the others, maybe the software shouldn’t react by pushing the nose down.

Didn't they have only 2 angle-of-attac sensors to begin with, and then decided, that only one would be fed into the software? So the software part had zero chance to even detect a problem with the sensor.