A man with one watch knows the time. A man with two is never sure.
In such a safety critical system, if you are using multiple sensors, you need an odd number so you can implement a voting system (and hope that there are no common-mode failures that afflict a majority of sensors, or at least engineer away the frequently occurring ones).
And, to be clear, MCAS, even when acting on incorrect information from the AoA sensors, never 'ran away' - it applied a one-shot adjustment which could easily be confused with normal running of STS. Unfortunately, adjusting trim with the pickle switches reset the MCAS trigger, allowing it to add another 'one-shot' adjustment, which if not fully recovered by the trim adjustment switches, added up with previous adjustments until the out-of-trim aerodynamic forces could not be overcome by use of the yoke. As history shows, this behaviour managed to fool several certified competent pilots who had not been fully informed of MCAS behaviour.
A continuous runaway is easy to recognise. An intermittent, slow piling up of small increments of unhelpful trim turns out not to have been obvious, especially in a busy work environment with falsely triggered alarms.