The methodologies usually adopted in water distribution system redesign problems consider the topology of the network as an input fixed datum; optimisation solely allowing for the duplication/substitution of existing components. In order to contribute to the identification of optimal solutions that may lead to a lower risk of failure to supply the required water, together with a lower redesign cost, this paper proposes a novel methodology which reports the influence of the existing network configuration and its performance. In particular, the redundancy of loops and the robustness of the network topology are investigated by applying an optimisation technique based on a genetic algorithm and by taking into account the random water demand at each node. The methodology presented has been applied to two case studies, in which it considers the influence of the topology on the overall system reliability/risk. The results demonstrate that it is possible to obtain further configurations that are more reliable for a lower redesign cost. The analysis performed highlights the impact of the topology on the search for an optimal solution, which, as a principal conclusion of the work, should be considered among the decision variables taken into account by the optimisation in a redesign problem.