a. Based on BEA factual report, that wasn't the case with AF447.
b. In such case, a few system design options to consider for autothrottle logic:
1. Use FMGC last known flight phase (climb, cruise, descent) as autothrottle reference to maintain optimal AOA, or
2. Set engine thrust used before the onset of airspeed indication error, or
3. Set CONT thrust.
The accident, onset by airspeed erratic readings (which lasted for one minute) and autoflight disconnect, the problem was lack of stall recovery by the pilot, for which he had 3:30 minutes and a mostly functional aircraft. A pilot descending 10,000 FPM with 107 KT indicated, should not expect a pitot system to generate correct dynamic pressure readings. He should get the aircraft under proper control firsdt, and only then carry out non-normal checklists.
As military flight academy commander, to me this case clearly shows the lack of proper selection (for personality and psycho-motoric skills) and training of pilots (complete flight envelope mastering in aerobatic trainers including deep stalls and spins, for better understanding of flight mechanics and human impact. Not necessarily to be used, but rather avoided. In case they are encountered, a pilot at any level should recognize and counteract out-of-controlled flight. It could also help him understand as captain, why full repeated pedal deflections in A300 might shear the vertical tail, what entering an aircraft wake feels like and how to avoid it. All is basic flying stuff). AF447 is just one of many such cases, including B737-800 in automatic ILS approach, throttles in idle, stalling before touchdown in Schipol, or a captain woken-up around FAF to watch his A320 demolish. Everything I wrote in this paragraph is 100% applicable to transports, even if many never thought about it.