To answer: I think I'm following this point here, but it leads me to questions. Suppose the sequence is yes-yes-no-yes instead of yes-yes-yes-yes? How does the dog differentiate the 3 things he did correctly from the one that he did wrong if the treat is deferred? Aren't you reinforcing the one mistake as much as you are reinforcing the 3 correct behaviors? Plus, how does this square with what seems to be a central point shared in nearly every clicker training instruction I have read? That being: it is essential to reward immediately after the correct behavior/click?
Yes-yes-NO = do over go no further. Yes-yes-close but lowsy form = do over go no further Yes-yes-yes-reward.
The click itself helps the dog differentiate correct-correct-no click must be wrong. The click becomes almost as strong as the treat. Well, maybe not quite as motivating, but certainly as information delivering. Click comes to mean YES and no click is NO. Just as treat = yes and no treat = no. Not as tasty, but certainly as much postive/negative feedback.
The very first exercise of owner-only dog class is t shape behavior with click/no click. Have your husband open the front door by hot-n-cold clicking him. Now apply to dogs to shape behavior. Our trainer said start by clicking in the room, click/treat/yes do that a bunch. click=yes=treat. Now do it with sit they already know. Oh yes, I DO something and I get a click=yes=treat. And down, and heel, and come, and stay. All the basics we know verbally and by hand signal from puppy class/basic. This part is the click/treat immediately bit. Do that for some time which REALLY cements the task=click=yes=treat. Now, use it to shape other tasks. I don't just want a sit. I want a call forward - sit - come around - sit - heel off. Just like memorizing text in school, start at the begining with clicks while successful, an "eh" or "nope" or "oops" at the wrong. You are not deferring treats for half an hour and the tasks are usually not that long. It does not take the dog but a few seconds to complete call forwar return right. Each part done successfully gets the click and the completed move gets the treat. Then when accomplished right fully you go back to a treat for the whole maneuver. call forward-return right = click/treat.
However, in the same vein, let's say I am going to run through 10 rally tasks set up with cones. The dog knows every task in isolation but needs to be fluid and quick putting them together (practicing). It's not about treating for every station which you can't do in the competition. You can click as you go through and the dog knows it is right at each station and no treat means there is more to do so keep focused. At the end of the run click/treat/good dog. If the dog mis-performs station 6, you re-try, if that still is missed move on with clicks for the rest. Afterwards, do station 6 over in isolation and then run through again.
There are levels to clicker use just as there are levels to training. The field lab being signaled with a whistle or hand signal to go left/right/forward/etc. You can't run up to treat. The border collie learning the herding while running away from you you can't run up to treat when it followed your signal. You have to be able to signal correctness at a distance without dealing with treating. You reward the lab when he brings back the game, the BC when he gets the flock in the pen. You have to have a yes signal along the way, something they know means right.
Maybe some of the dog trainers here can explain it better. I am only going on how it was explained by our trainer and how it has worked successfully in shaping my dogs and the dogs in my class into doing multiple step tasks.