tdracer;
Thanks for the elaboration on your meaning of "deep stall". Davies discusses the phenomena in Handling the Big Jets - what you've described re the BAC111 is how he describes it.
AF447 was descending a bit faster than it was going forward, (12,000fpm, or about 130kts) but although reduced, the elevator still had authority.
While the stick was never held in the ND position for "10 seconds", there is a time period when it oscillates between NU & ND but is mainly ND for about 10" at around 17000ft in the descent and the pitch can be seen to change to ND and the AoA moves from about 43deg to about 38deg; then the stick returns to NU, essentially until impact.
While the sim does not have stall algorithms from actual flights past the approach-to-stall, going from memory the sim took even longer than the time you've estimated, (~45" from beginning of recovery at around FL350 to recovery at FL220, approx). The stick was held in the full ND position for the entire time, the descent rate reached 20,000fpm. The flight path vector symbol moved up very slowly from it's initial indication of ~ -35deg; the THS began moving from its full -13deg (NU, tail down) position a few seconds after the stick was held in ND position - the pitch was controllable throughout, though very slow to respond.
Recovery is 30" away...
As you know, air is thicker lower down and control may have been better and recovery time a bit shorter. I'll bet they're worked on algorithms to simulate an actual full stall. Still, the sim was pretty good - shake, rattle and hum, with the "Stall, stall, stall" message blaring - oddly it never quit until the sim became unstalled so not sure the NCD loss of the warning situation is reproduced.
One last item - I had thought the same thing regarding the engines - that high incidence angles would upset the engine to the point of putting the fire out but they both remained operational (and at high power) throughout, until impact.