It is well recognized that a single, arbitrarily efficient solver can be significantly outperformed by a portfolio solver exploiting a combination of possibly slower on-average different solvers. Despite the success of portfolio solvers within the context of solving competitions, they are rarely used in practice. In this paper we give an overview of the main limitations that hinder the practical adoption and development of portfolio solvers within the Constraint Programming (CP) paradigm, discussing also possible ways to overcome them and potential extensions outside the CP field.