In this discussion I will propose an agent architecture for the representation and simulation of agents working at different time granularities. The main point of this talk is to show a policy for resource distribution which takes advantages of the potential influence one agent may has upon other agents of its own class and other classes. The example I will use as illustration is a simulation of a forest, but we may also discuss about other kind of agents in population dynamics. Important notions to be presented are that the resulting system is free of dead lock and starvation (in computational sense :-)).