To put it another way: on the face of it, it seems bizarre for the FBW to "grab" a transient, higher-than-normal AoA even if it results in more than 1G with a neutral sidestick (assuming wings level). No doubt there are good reasons for doing so - perhaps the likes of OwainGlyndwr may explain?
Sorry Chris, but the likes of Owain Glyndwr have no explanation other than the obvious; that the logic designed to protect against stall at low Mach numbers had unexpected side effects when applied at high Mach numbers.
The low speed logic is not stupid. If the system has reason to expect that stall AoA might be reached in a short time if current trends continue then it puts the aircraft into a temporary protection mode. If the pilot genuinely wants more AoA then it is available by moving the sidestick back. Otherwise the protection can be removed and the aircraft returned to "normal" behaviour by moving the stick forward. [Looking carefully at a digitised version of the traces it looks as if the "short time" is about 1.5 seconds]
In this particular case the autopilot dropped out because of a strong temperature shear (not the windshear you suggest). I personally suspect that designers do not currently give enough attention to temperature shifts, perhaps because they are not specified in FARs or CS25. On Concorde, which was designed against its own special (and generally more severe) rule book, we had to consider temperature changes of up to 21 deg C in one mile. This A340 incident was only 10 deg C in that distance.
Just after the A/P dropped out the aircraft met some reasonably strong turbulence. Again not extreme; working from the normal acceleration record one might deduce gusts of up to 15 fps (EAS) which is just under half the design gust for those altitudes. However, some of those gusts were sharp edged, so the rate of change of AoA was high. The actual AoA remained well below the deterrent buffet stall level - round about 2 deg below the Vs1g level, but the predicted AoA for 1.5 secs later went over the stall limit and triggered the alphaprotect mode.
Accepting that the subsequent motion was unexpected, it was not IMHO intrinsically unsafe. This was not an embryo AF447. The EFCS was, after all, acting to prevent a stall. There were no injuries and no aircraft damage. Dare I say that were it not for the proximity of the A330 and consequent possibility of a midair collision this incident would not have received the attention it has had. That seems to me an ATC problem and I note that the AIB made two ATC recommendations but nothing on the airplane.