The most likely scenario in my head is some sort of control surface malfunction. Given how quickly the debris was located I'm pretty confident that they'll get the black box and solve this mystery.
Yeah, a quick (and very elementary) back-of-the-envelope energy calculation shows that gravity alone couldn't achieve that speed. [SEE EDIT]
> Were the pilots desperately trying to go up, not knowing which way was up?
We'll have to wait and see, but it was morning, and I think I heard that there was weather in the vicinity of the crash but not at its actual location.
EDIT: That calculation was wrong. The aircraft had more energy at 5475 ft than it did at 3200, 7.73 seconds before the last data point.
As difficult as it is to believe for anyone who received their training in almost any other program, the crew of AF447 had never practiced a single stall recovery in an aircraft before that flight.
In addition, the transcript shows that they were not mentally frozen in a panic and unwilling to act, but that they were poorly-trained to correctly identify the problem, something compounded by terrible CRM as they argued about it such that even those who had correctly diagnosed the problem were overruled after the time to test their hypothesis had been wasted in argument.
Instrumentation issues also compounded the above, among other things.
Lastly, the pilot who attempted to apply correct control inputs to recover had his inputs negated by a dual-control input system so mind-boggling that it could only have been conceived by a company with such a culture of arrogance-driven ignorance as Airbus.
Would you like to help me out with that?
If all potential energy from an altitude of 5000 ft = 1500 m got converted into velocity, we'd talk about 170 m/s or about 330 knots (and a free fall from that altitude would take about 17 seconds, t = sqrt(2h/g)).