Sheldon said crappy QRs and no lawyer lips would lead to disc bikes ejecting their wheels upon braking.
I thought about this, and then (after running through a bunch of simulations in my head, before figuring out the relevant geometry) figured out that the placement of the brake itself affects this.
When braking, the brake becomes a pivot point for rotation. Let's say you're looking at the bike such that it's traveling left. The wheel turns counter-clockwise going forward, and will always do so. So if you brake, say, up and right from the axle, then the axle will go mainly down and forward--attempting to jump out.
This is where they put the brake.
It's easy to mount forcefully to the fork there, sure; the engineering problem of putting the brake elsewhere is difficult. For example, up and left is a tricky place to mount; but will pull the wheel up and back--into the drop-out--during braking. The downside, of course, is that the counter-force will pull the brake mechanism away from the fork; whereas where it's currently mounted will pull it toward the fork.
For that, I say to mount a bracket around the brake, which goes back behind the fork and mounts ... exactly where the current mount points are. This means the brake pulls forward against the bracket and ultimately into the fork, rather than away from the fork in an attempt to rip out the mounting screws. Of course, that means you have a bunch of crap in the way--a mounting bracket that can interfere with cables or brake mechanisms, depending on how it's mounted. There are many ways to get around this, not many are elegant.
Still, positioning the brake strategically would lead to forcing the wheel into the frame, rather than ejecting it from the drop-out. The rear brake... I think is mostly fine, as it pushes the wheel forward and the drop-out points back(?) (probably because its failure mode is to drive the wheel into the frame as it drives forward). Somewhere in the lower left quadrant is probably best.
None of this matters, as changing any of this would create large incompatibilities with current frame and brake design. You could technically keep the same mounts and mess with brackets up front; that's probably a good idea to avoid a rip-out style failure. Still, the exact mounting geometry for optimal braking performance may turn out slightly different, and so you get compatible-but-not-great versus great-but-let's-break-everything-again. Going the latter route with a bracket to adapt current forks to take the changes is probably the best way, as it gives a way to sidestep the incompatibility issues....
... but of course there's a large cost associated, and you create a mixed market for a while.