I've been dinking around with this and here is what I've come up with on the right turn after the stall (after all, why not a left turn?)
First, there was no "uncovered setting of the rudder trim that became exposed by Alt2 as I had earlier suggested there might be.
The autopilot has authority to trim the rudder in the same way that the pilots can and did so a number of times up to the AP disconnect. It was with the last A/P rudder trim setting that AF447 did its descent.
Before the AP disconnect, there are some small trim adjustments on the order of 0.2 degrees. ranging from -0.25 to -0.5 degrees. After the AP disconnect, there are some small blips in the trace, but the indications remains as before. The position moved to -0.4 degrees seconds before the AP disconnect and that is where it stayed. -0.4 degrees has the effect of being a slight
right rudder input. The reasons for the blips in the trace are unknown, and are most likely caused by noise on the circuit, or mechanical slop in the indicating system.
The effect of slight rudder trim would be minimal until the AOA began to run up, at which point, it would cause a preference to drop the right wing in a stall. The turns that are observed after the stall would then result from this preference plus pilot inputs on the rudder.
This probably repeats the evaluations of others, but now is my own.
When I pulled the clean F-4 into a stall and held it there, the aircraft would drop off on a wing. If I then applied opposite rudder gradually increasing, it would then flop over on the other side without being able to stabilize it even momentarily in a wings level mode. If I continued doing this ruder exercise, the aircraft would reverse itself yet again ever more violently. This is the experience I use to explain the AF447 post stall behavior although the A330 has a very docile stall by comparison.