I think VicMel's analysis is compelling. The AoA vanes are probably conforming to spec and the fault is further down the chain. It is possible that there is a fault in the nose to wing conversions as described. But these faults are reasonably easy to prevent through peer review processes such as code walkthrough and also module testing. What is more difficult is if an event causes some other subsystem to write garbage into the working space of the AoA handler, and that event is low probability. This sort of thing is really difficult to find in real-time systems.