If we get data off the CVR, then I have a strong suspicion we are going to find long periods of silence leading up to the loss of control.
Falling asleep on the job is not the same "black and white" process we normally experience each night. When we fall asleep normally, there is a gradual but one way transition from awake to asleep, and once asleep you stay asleep. This is because your brain is not sending contradictory signals to stay awake.
When falling asleep while trying to stay awake, the process is usually a lot more chaotic. Brief microsleeps and long periods of being "half-asleep" (where you are technically awake but have very low levels of awareness and responsiveness) are interrupted by sudden periods of wakefulness as you catch yourself falling asleep. Critically, judgement is severely impaired during this process and the individual has very limited insight into their state of mind.
During this process however, it is quite possible for some intermittent purposeful activity to take place. A driver will navigate a corner in the road. A pilot might try and make a radio call. Sooner or later however, a microsleep will coincide with a bend in the road.....
I believe this process explains the failure to divert around weather, and also some of the communication inconsistencies. However the flight continued to proceed uneventfully as far as the Captain (? sleeping in the rest area), the CC and the pax were concerned. Suddenly something major happened (? initiated by pitot tubes freezing, ? weather related upset), and it all turned to s.....
Its not just pilots. ATC also.
Fourth ATC Sleeping Incident At Reno | AVIATION WEEK
On the surface, it is staggering that in 2011 commercial flights can land at the nations capitol with ATC asleep at the screen. Given the recent spate of sleeping ATC in US, I wonder if a similar problem is not the explanation for the apparent staggering failures of ATC that were responsible for the passage of AF447.
If this is the explanation, we need to look at the system issues rather than point the finger at the individuals concerned. Humans are not bad working through the night when the job involves either manual work, unpredictability, or continuous cognitive processing. Police officers on patrol or a cab driver in a busy metro area looking for a fare are not likely to go to sleep. It is more a problem with routine activities (such as long distance driving), or where the job require oversight and monitoring in case of an unexpected (and unlikely) problem.