I think you are overstating the type of communication that goes on during dynamic play. There are protocols on what the ARs will say. I know for MLS that the AR protocol is that if the AR has OS it is the AR that makes the decision on delaying the flag, and the full story is communicated by the AR saying "Delay, delay, delay" into the comm. The AR then raises the flag when the AR concludes that the attack no longer meets the criteria for delay. (And of course, as with all flags, it is up to the R to "accept" a flag from the AR.) I would be surprised if the PL logistics are much difference and include the R advising the AR whether to delay the flag or not.
My sense is that MLS has done a much better job than PL of training ARs on the delay practice and creating a consistency in application. The delay process is a very hard one to implement as it busts up years/decades of muscle memory and brain process. Now, on every OS call, the AR has to also process (1) was this close enough to delay (when we have always thought of these as black and white, now the AR has to think in gray) and (2) is this a "very clear attacking situation" (a brand new, not crisply defined concept)--all while at the same time being prepared for a potential next call