Originally Posted by
CONF iture
In the meantime, if the aircraft had clearly announced what it positively knew "My AoA probes disagree", the pilots of the day would have never proceeded further with such a test.
That may be true. But airliner annunciations are not, in general, designed for test flights; they are designed for normal operations, and to minimize workload by masking messages which are not expected to be critical - so called "nuisance" or "status" messages. In this case the system thought it was rejecting the one bad value and still had dual redundancy - plenty of margin of safety for a flight where stall approach is an unlikely event. But in this case such a manoeuvre was probability=1. Which throws all the usual risk calculations out the window.
Again, this is why when conducting tests there are different procedures to follow. Perhaps (a/c dependent) you need to check status pages to confirm no faults, or be alert for messaging that would normally be dismissed, or for characteristics that are usually benign but not in these circumstances.
That the ADIRUs didn't deduce from the stuck values that the AOAs were erroneous but valid is unfortunate, but "erroneous but valid" - especially from multiple independent (even if similar) sources - is a horrible failure case to defend against.