A mess indeed... but look at the Qantas case, the preliminary report if I understand correctly reveal some potential problem linkage between the ADR and IR part of the boxes... which needs to be looked at....
PJ2,
The qantas case might be explained from IR1 providing wrong data to the calculations of the ADR1 output. AOA, calculated airspeed, etc all probably have an inertial component in their calculation/correction.
The opposite (an IR1 fault beginning with an ADR1) is apparently much harder to demonstrate.
What puzzles me the most in this AF447 case is that IR2 fault reporting...This is what is making me refuse to accept that the problem came - at least exclusively - from the pitot/static system.
It is not logical to have inertial data validated by air data - hence the name, inertial.