bear/dozy:
If I may, we need to be careful and aware as we discuss this regarding the difference between what speed was through the airmass, and what speed was indicated, since an axiom in this case is that the sensed and processed airspeed was (for at least part of the event chain) unreliable, and not a reflection of the actual speed through the airmass.
This got me to thinking. Will BEA be able to compare inertial references to isolate when true airspeed was being detected before things went awry. You could backwards compute it from ground track for x minutes before the event. You would come up with wind and true airspeed then, and maybe get a better handle on (by using ground track compared to Airspeed during the event) a finer estimate of what the plane was doing despite what it was displaying ... airmass wise.
Hopeful.