This will IMHO turn out to be a failure of FMECA -
https://en.wikipedia.org/wiki/Failur...ality_analysis. The programmers failed to test the case where the AOA sensors disagreed and then project forward to the effect on the aircraft if the pilot either didn't know about or was slow to disable the autotrim. Code should have been written to either limit the horizontal stabilizer travel or to find another way to confirm that the angle of attack was really nearing the stall angle.
OTOH, if it is really necessary to drive the trim to full nose down to prevent stalls then arguably there is a fundamental design flaw in the aircraft. Maybe it just needs bigger elevators period.