Simply falling on this switch wouldnt cause it to change positions - it requires a deliberate act to do so, the switch requires a certain force to pull up and then move to one position or another, its not like accidentally changing channels on your TV because you sat on the remote.
I can believe this. But what if, instead of falling against the switch, the copilot, recognizing that he was about to pass out (e.g. recognizing symptoms of an impending stroke), intentionally attempted to move the switch to the "unlocked" postion (to make it easier for the captain to get into the cockpit quickly)? Due to a combination of confusion, physical incapacitation, and infamiliarity with a probably rarely-used control, he could conceivably have turned the switch to the wrong position even while he was attempting to do what he thought would be the best possible action.
Also, there is no button or switch he could have fallen on which would have caused the gradual descent that we know the aircraft took. Changing the auto pilot altimeter requires you to use a dial and then confirm the change in two separate actions. Any interaction with the side stick would require the auto pilot to be off, which would mean we should have seen a lot of other, large movements in the aircrafts path, which are completely missing from the telemetry we have at the moment.
The few commands that we see in the telemetry (and by telemetry I mean the transponder tracks, which cover speed, height and directional changes) indicate that the aircraft was under either the control of the pilot or the autopilot for the entire duration of the descent.
As far as I can tell, this is nonsense. Under "normal law" in an Airbus autopilot system, sufficient pressure on the control stick will override the autopilot system. For downward pitch, the autopilot system will allow up to 15 degrees of downward pitch to be commanded without removing the autopilot system from "normal law"; other autopilot functions will continue to function normally. I'm not sure what the exact result of 15 degrees of downward pitch would be, but I'm pretty sure it would be a rapid but controlled descent—exactly what the telemetry shows.