It's fair to say, I think, that implementing synthetic airspeed is a far greater effort - development resource wise as well as computational resource wise - then validating two AOA readings against each other.
Given the time it took them (or better takes them, 'cause their are still not done) and the problems they encountered when implementing the later gives a pretty clear idea of a lower limit time and trouble wise synthetic airspeed is going to take. Do not forget that the synthetic airspeed information has yet to be included in other systems, their diagnostic routines, mode switching, warnings, etc.
Personally I would expect either upgraded hardware or maybe even another box (e.g. 3rd air data unit), which might be a more effective solution, becoming neccessary. Kalman filters are efficient but have tight real time requirements. From what we have learnt about that FCU ... difficult.