Having read most posts since the recovery of the recorders, and coming from a medical-statistical background, DozyWannabe in #833 raises an important issue:
What we have here is an edge case where that design decision may not have helped matters. What we don't know is the number of times where that design decision may have helped in a situation where the flight envelope was not compromised in such an extreme manner.
When checking for errors/illnesses, the false negative type is typically the one that gets most attention, while false positive alarms may be more harmful in the long run.
False negative in this case would mean not to report a stall, that is indeed happening, while false positive would be to report one that isn't there. A false positive alarm is a cry wolf, insensitizing pilots to a particular warning. The "shut up gringo" incident, the tripped circuit braker on NW255 and (maybe) AF447 reactions are indications of this happening.
Actual stalls are very rare, so even though a system may have a low propabilty of reporting false positive, that scenario is still more likely than a real stall (true positive). In general, minimizing both types of errors may not be so simple, and lowering one type can lead to a rise of the other one.