I can't say anything rigorous, but it would probably drop the number of gm level blunders drastically.
Combined ELO of more than one person?

I suppose you can envision a Venn diagram, with each circle representing the set of moves each player not only sees but also understands (I, for example, may see a perfect move, yet not recognize it as being perfect, and may pass it by in favor of an inferior move). The better the player, the larger the circle, but there is only so much expansion that can take place. Combine two great players and you would reduce the chances that a great move will be missed; the two circles overlap considerably, yet still have areas in which they do not overlap, so value is still being added. A player like me would have my circle lie entirely in that of a better player, so I would not be able to help him or her at all. Likewise, just as 20 years of experience may in fact be nothing more than one year of experience 20 times, a bunch of average players will probably also have tremendous overlap in their circles, and none will see the really deep theory that suggests certain moves. So there will be a point of diminishing returns. But two or three will be better than one.

I think the best advantage of it is the declining of blunders. A lot of times blunders are due to a moment of blindness, and I think most of the time one person in the group won't be totally blind to some obvious move that another player in the group maybe missed in his calculations.
More players also would give you access to more styles of play. For example, a 2400 player might be 2400 because of his exceptional tactical ability; another 2400 might not be quite as good at tactics, but has a fantastic positional understanding. So you can sort of get the best of both worlds there. It's probably a decent effect -- maybe 100 points or even more in the best of cases -- but like you said, the effect increases less and less as more players are added. I think no matter how many 2400s you have, you still can't understand the game like a 2800.
My question is prompted by a comment at the recent London Chess Classic in the initial press conference where the director said the combined ELO points of the assembled competitors was 27432. Of course, you can't just add them like that. So I was wondering... if you have two players of 2400 rating (say) working together to choose moves, what would their effective ELO be? My guess is maybe 2500 if they are communicating well - but this is just a guess.
Does anyone know if this has been investigated more rigorously? Is the ELO boost gained by having two GMs work together more than that gained by two club players, or does it depend on the absolute level as well? Presumably the gain gets less and less as you add more people to the 'team', otherwise Kasparov could not have beaten 'the world' in 1999.