Rust hole in chain stay bridge

I'm sorry but I have to disagree with your hypothesis.

I can see why it appears to make sense, with a forward facing horizontal dropout (ffhd) the upper run of chain pulls the axle up into the top of the dropout as you drive the pedals whereas when you resist them the lower run of chain applies its force inline (or at least much more so) with the slot in the dropout so all that's holding the axle in place is the nut/locknut clamping the top and bottom of the slot.

Where I think this falls down is when you consider the track end. In it's usual application on a track or fixed wheel road bicycle the chain applies its forces at an angle that is approximately in between the two scenarios in the ffhd and that angle is much the same whether you are looking at the top or bottom run of chain. Wheel pulling is of course not unknown with track ends but, in my experience at least, is much more common under driving than braking (I'd go as far as to say almost unheard of under braking) and usually there is more at play than a lack of torque in the wheelnut.

As previously mentioned, the angle the force is applied at is the same for driving and braking with a track end so if wheel slippage is more common under drive then something else is at play. I'd guess that the average rider is unable to push back with as much force as they drive forward and I'd also guess that the momentum of the bike moving forward keeps the wheel which is being decelerated from sliding forward in the track end.

If we now consider less common applications of the track end, bikes used for polo, speedway, artistic cyling and radball, we would find two common denominators between these applications. Smaller chainring sizes and lower gear ratios. Smaller chainrings will apply the forces to the axle/wheel much more inline with the slot in the end (akin to how it is applied when resisting the pedals in a ffhd) and the lower gear ratios would mean more torque (not sure if torque is quite correct but I'm sure you know what I mean) which would be more likely to cause a slippage. If a wheel can be adequately retained in a track end on these types of bikes then there's no reason why it cannot be adequately retained in a ffhd.

Where I know a ffhd does fall down is that they are often thinner than track ends and other modern dropout designs and as such the stub of a qr axle which protrudes beyond the lock nut can be longer than the dropout is thick so the qr skewer bottoms out on the axle before applying the proper clamping force to the dropout. This leads to the myth that it is impossible to hold a wheel in place in a ffhd.