I think all of the issues that make the collision model such a challenge are also in play anytime you "join" someone.
I think if you're riding along and viewing a 1v1, in actuality there has to be 3 "versions" of reality at play. The two participants don't see exactly the same thing at the same time, and I don't see how the joiner can see what either of the participants see.
Due to lag alone, even if you saw the things you wanted to see, you'd never see them at the right "time". You may see your host pilot fire, but it would always appear to be "off", because neither he nor the target are where you think they are, when you think they are, hehe. I think it could even lead to learning the "wrong" time to fire, drop/raise flaps, manage throttle, etc.
Ride alongs work best (IMO) as a way to learn SA, how to approach fights, etc. More of the "vague" things that can be tough to convey. Things like threat assessment, who to fight first based on threat, how to judge relative E-states, and even things like whether this particular fight needs to stay fast and open vs slow and tight based on surrounding elements. It'll show things like fighting in the vertical, specific manuevers etc, it just won't show the true timing of those in relation to the opponent. Close enough to probably get the message across, but not close enough for finer aspects such as gunnery, etc.
I also like it to see if help people get a grasp of where to look and when, watching for everyone nearby. They can work on that aspect of SA without having to manage the plane, and decide where the fight needs to go, etc...
I think films are probably ther best/only way to really "see" what another player sees, in his/her version of "real-time".
MtnMan