Asraam
Exactly HOW does your stick pusher know it is approaching the stall, perhaps it will use the AOA probes ... that generated the reversionary mode in the first place.
By reverting to the inertial reference system (IRS).
I can understand why AB would not want to use IRS data for normal law, to define a stall, it could easily be used for alternate law.
The IRS knows the attitude (after all, that is what we fly by) and it also has a good idea about the airspeed.** And the FMC knows the weight. From this, the correct stall attitude can be calculated with reasonable precision.
** The IRS and FMC know the groundspeed, while FMC knows the previous windspeed and altitude from before a problem occurred, and can use this to calculate a probable TAS airspeed. The FMC could also integrate known thrust, attitude and altitude, to derive a secondary airspeed profile.