The concept is simple enough: no pressure in the train line has the brakes locked, full pressure is loose, and partial pressure drags accordingly. So if pressure is lost for any reason, the train stops. (or at least it's supposed to; more on that later) So I'd expect to see the brakes applied by a spring that is capable of stopping the wheels of a fully loaded car, and a single-acting cylinder that works against that spring to release the brakes. A hand-brake would also work against the spring, in parallel with the air-cylinder.
But everything I've seen to explain how it actually works includes the same seemingly unnecessary complexity, as just a matter of fact with no mention whatsoever about why. They do, however, mention the pitfalls of this overly complex system and how it's caused accidents.
Likewise, every car that I've had access to crawl under seems to confirm the explanations. The obvious direction of forces (thin pull rods for a direct giveaway) requires positive pressure to apply, not positive pressure to release.
The way that it actually works, has a local reservoir on each car that is "charged" by the train line. When the train line reduces pressure, a purpose-designed valve allows a corresponding amount of pressure from the local reservoir into the brake cylinder. This positive pressure applies the brakes.
There is also an emergency function that senses a sudden drop in pressure and finishes dumping the train line at that car. Then the next car sees a sudden drop, etc., thus propagating an emergency brake signal much faster than would otherwise be possible.
The obvious pitfall that has indeed caused accidents, is that it takes time to recharge all of those local reservoirs. So if you've just released the brakes, the cylinders are dumped quickly, but it's impossible to reapply them for a while, at least to full effect, because the local reservoirs haven't recovered yet. Then I see more complexity being added, in the form of a second train line that is always "charged", for the purpose of keeping the local reservoirs topped off...
The only reason I can think of to not use a direct system, with a spring to apply and direct air pressure to release, is to reduce the time from completely-locked to completely-loose, as the local brake cylinder can be dumped to atmosphere effectively instantly. The emergency function wouldn't be affected, since it can dump the cylinder to stop in the same way that it presently dumps the cylinder to go. And there are "fast-return" valves for "ordinary" pneumatic actuators that dump the difference in pressure to the atmosphere, which in this application would allow the brakes to be applied with no more delay than they are presently. (the train line still only needs to discharge its own volume, not the cylinders') At this point, the only remaining difference is that it takes longer to release the brakes, as the train line is filling the cylinders instead of reservoirs.
It may sound like I'm adding the same level of complexity as what the present system has, but I think a key difference is that none of it is absolutely required for the system to work at all. It's only an enhancement of what's already there, compared to a fundamental change in operation if part of the present system is removed.
Is it really worth the complexity and risk to have a train leave that quickly? Or is there something else that I'm still missing? To use some electronics terms, why is it necessary to reverse the polarity of the control signal on each car and rely on a small amount of stored energy that is slow to replace, instead of using an actuator that takes the supplied polarity directly?