Haha this is a fantastic answer. Also, not most of the effect, all of the effect.

Do the same experiment with the paper, but rest the trailing edge on a table. The air can no longer be deflected downwards, so the paper will not rise.

Another one to think about, at the air metal boundary the air is stationary on both sides of the wing. So why is there a pressure difference? Bernoulli's law is valid only along flow lines and is a consequence pressure difference required to accelerate (or decelerate) a fluid. Since in the laminar situation flow does not occur from the top surface to the bottom (minus around the tips), direct application is nonsensical. If you integrated all flow lines from all surfaces however, you would get the correct answer.

Well, scale up the wind according to the weight of the metal (and the overall surface area of the metal) and it can happen, as evidenced by this 747 in a boneyard 'rotating' under heavy gusty winds [0] (taking into account the massive weight and CoG difference without the engines on).

Extend the flaps and slats on that aircraft and I am sure the effect will be all the more pronounced.

And yes I've flown aircraft inverted too - it take a tremendous amount of forward stick to try and maintain level flight in that config to counteract the wings natural tendency to move towards the top surface.

People have pointed out the "Thunderbirds" F-16s flying in the "mirror" formation [1] and not displaying much difference in AoA between the upright and inverted aircraft, but supersonic fighter usually have a straighter, almost trapezoidal shaped wing cross section rather than a curved 'fish' shape. I am willing to bet the inverted pilot has a fair bit of forward stick on though.

[0] - https://www.youtube.com/watch?v=cHhZwvdRR5c

[1] - http://c8.alamy.com/comp/EG175E/little-rock-air-force-base-a...