It sounds to me not just a software issue but a systems issue where software, hardware and architecture play a role. What used to be a sensor to measure the angle of attack became suddenly key input, single source in a short term control loop for the most vital axis. What could go wrong...
Taking into account the other sensor data is an obvious choice. A better understanding the quality of the sensor data in real field operation may also be required. Clearer stabilization problem indication, quicker manual disengage of auto pilot (several reports of nose down events) and MCAS are others. Decreasing the allowed tail trim for MCAS is another.
Then that leaves: Another one or two sensors. Independent review of the engine control laws (report of unexpected low thrust when climbing).
There is a lot to be done. The plane was rolled out too early and fixing the cut corners for real could take a while.
Taking into account the other sensor data is an obvious choice. A better understanding the quality of the sensor data in real field operation may also be required. Clearer stabilization problem indication, quicker manual disengage of auto pilot (several reports of nose down events) and MCAS are others. Decreasing the allowed tail trim for MCAS is another.
Then that leaves: Another one or two sensors. Independent review of the engine control laws (report of unexpected low thrust when climbing).
There is a lot to be done. The plane was rolled out too early and fixing the cut corners for real could take a while.