The nonlinear constraint routine may be called multiple times before finding a point within bounds to be run through the objective function. However, the optimizer breaks up the problem into multiple sub-problems that it believes / hopes will not violate the nonlinear constraints. Therefore there is not a one-to-one relationship between f(x) and c(x) being called.
If you have an expensive function that must be called by both, then you may wish to implement that function with a "cache". See for example http://www.mathworks.com/matlabcentral/fileexchange/49949-cachedcall . A similar version that saves in memory could also be programmed.
