In this paper we consider the vehicle routing problem with stochastic demands (VRPSD). We consider that customer demands are only revealed when a vehicle arrives at customer locations. Failures occur whenever the residual capacity of the vehicle is insufficient to serve the observed demand of a customer. Such failures entail that recourse actions be taken to recover route feasibility. These recourse actions usually take the form of return trips to the depot, which can be either done in a reactive or proactive fashion. Over the years, there have been various policies defined to perform these recourse actions in either a static or a dynamic setting. In the present paper, we propose policies that better reflect the fixed operational rules that can be observed in practice and that also enable implementing preventive recourse actions. We define the considered operational rules and show how, for a planned route, these operational rules can be implemented using a fixed threshold-based policy to govern the recourse actions. An exact solution algorithm is developed to solve the VRPSD under the considered policies. Finally, we conduct an extensive computational study, which shows that significantly better solutions can be obtained when using the proposed policies compared with solving the problem under the classic recourse definition.