It is quite simple, and said before.
In high AOA and especially in the stall region the aileron is the wrong control surface for roll control.
If you like, use any aircraft model to get a picture from following description.
In normal LAW by a roll action on the side stick, the pilot commands a roll rate, with minimized sideslip (automatic turn coordination), elaborated in the FCPC.
In Alternate 2 LAW the roll control is direct, an order on the side stick directly commands a deflection, according to a kinematic. The yaw control is achieved from the pedals, through a mechanical linkage without automatic sisdeslip correction. A Dutch roll damping function using limited yaw rate data is introduced via the yaw damper servoactuators.
When PF wanted to raise the right low wing with a left SS input, the downgoing aileron on the right wing produced a lot of drag, thus inducing a yawing momentum to the right (into the low wing).
Normally the upgoing spoiler on the left wing would not only help in roll, but would produce drag as well and thus reduce the yawing moment, but under such high AOA the spoiler is blanked out, does not assist in roll and produces not enough drag to prevent the yaw. Sideslip protection was not available, and manual rudder not used.
The induced yaw into the right (low) wing increased the sweep back in relation to the relative wind and thus decreased the available lift on this wing, where on the opposite wing the sweep in relation to the relative wind decreased, thus increasing the lift.
Why did the bank not incrase beyond the 40° (like going into a spiral dive) but oscilated between some left bank and lot of right bank? I donīt know for sure, but it might have something to do with the dutch roll dampening function:
In the event of loss of the inertial data from the ADIRUs, the yaw rate data for Dutch roll damping are provided to the FCPCs via a unit comprising 2 rate gyros. In case of 3 FCPCs loss, the Dutch roll damping is achieved by the FCSCs with yaw rate data coming from the rate gyro unit
Thus left SS input without rudder input produced exactly the opposite what the PF intended to achieve.
One more comment concerning the stall behaviour. It looks like most pilots expect violent maneuvers of the aircraft in a stalled situation, heavy g variations, fast nose position transitions like bank pitch and yaw changes or what so ever. I expierienced the stall violent at the point of departure, but after those initial unpredictable movements nearly stable, but just not logic responsive to the flight controls. AF447 departed and behaved similar, and in combination with a different expectation might also have led the crew to take a stalled situation not into consideration.