Suppose there is an area to survey. If we deploy one robot it takes time T to sweep the area. Now if we deploy 2 robots, we would expect it to take time T/2. However in reality, because we are sending two robots, we would have to implement additional coordination planning routines and obstacle avoidance routines so that they do not collide with each other. These overhead computational costs add up and we will actually see the area being surveyed in time T/2+X. Now could you suggest papers where they talk about the estimation of X? It does not have to be in the context of area surveying. It can be any other multi-robot task also.
I came across one paper that had a good description of this issue from Kristina Lerman and Aram Galstyan at University of Southern California entitled "Mathematical Model of Foraging in a Group of Robots: Effect of Interference"
I think the question relies too much on variable factors so that there is no definitive answer. In a worst case scenario, you want to map a tunnel system and all of your robots start at the single entrance to the system. All of your robots will have to drive together to the first intersection where they can spread. More robots will could decrease the performance of your fleet. In a best case, you have some swimming robots with sonars (so with limited sensor range) that should map the sea floor. If you don't have too many robots, the fleet performance will scale linearly as the boats are able to move immediately to unscanned areas.
Perhaps, the following related work can be somehow enlighting or provide useful insights, as per your request: