igh,
As I said, I think you're trying to make a distinction without a difference.
A 'true' or 'real' yaw-roll decoupling would mean to me a 'model based controller', i.e. A controller which incorporates a numeric model of the aircraft flight mechanics characteristics. That means the controller 'knows' in advance how the A/C 'will' react to a certain rudder deflection and thus commands the appropriate amount of aileron deflection 'simultaneously' and 'instantly' to fully compress the yaw-roll coupling induced roll. Depending on the actual implementation and the actuator rate capacity you wouldn't see any roll then.
1. An ELAC and any FCC in any FBW airplane is run by a bunch of code. I don't know what you want to mean by "incorporate a numerical model" in such a way that other FCCs have it but ELACs don't. They all have numerical calculations built in, and those numerical calculations are obviously based upon the numerical flight characteristics of the airplane, otherwise how could you get those wonderful Bode and Nyquist plots?
2. Controllers don't "know" anything. They run code. No FBW controller I know controls the airplane according to internal parameters alone (thank heavens!). All such systems are feedback control, which means that
* there are target values for parameters
* actual values of those parameters are obtained by sensing
* the controller issues actuation commands to bring the actual values closer to the target values
Now, this all happens at the speed of electrical signals through copper wiring, which some significant fraction of the speed of light, and the signals only have to go a few meters.
3. Nothing happens in advance in a feedback controller. And neither is there any significant latency in the digital control. Sensing, computing and commanding happen to all intents and purposes instantaneously. Sensing in unstable airplanes such as the F-16 happens at about 200Hz; I don't know the sensing rate for roll on the A320. There may be some latency due to the physical characteristics of the actuators, but this has little to do with how the controller is designed (except, of course, that such latency will be accomodated where needed).
4. Nobody is suggesting the decoupling is completely, rigorously effective in all situations. Indeed, posts from Chris Scott and others say most clearly that it isn't. But I don't see that the strength of the decoupling has anything to do with "back door" or "front door". The kit is clearly designed to decouple roll from yaw, as other posts (for example from Clandestino) show, and as you seem to agree.
5. I doubt decoupling is properly describable as a "side effect" of the control SW design in the ELAC, as you suggested. Decoupling is, obviously, designed in and is there quite intentionally. It's in the manuals, as we have seen. As are its limitations.
PBL