What if we increase the number of objectives? Theoretical and empirical implications for many-objective combinatorial optimization