We consider a routing problem in the context of large scale networks with uncontrolled dynamics. A case of uncontrolled dynamics that has been studied extensively is that of mobile nodes, as this is typically the case in cellular and mobile ad-hoc networks. In this paper however we study routing in the presence of a different type of dynamics: nodes do not move, but instead switch between active and inactive states at random times. Our interest in this case is motivated by the behavior of sensor nodes powered by renewable sources, such as solar cells or ambient vibrations. In this paper we formalize the corresponding routing problem as a problem of constructing suitably constrained random walks on random dynamic graphs. We argue that these random walks should be designed so that their resulting invariant distribution achieves a certain load balancing property, and we give simple distributed algorithms to compute the local parameters for the random walks that achieve the sought behavior. A truly novel feature of our formulation is that the algorithms we obtain are able to route messages along all possible routes between a source and a destination node, without performing explicit route discovery/repair computations, and without maintaining explicit state information about available routes at the nodes. To the best of our knowledge, these are the first algorithms that achieve true multipath routing (in a statistical sense), at the complexity of simple stateless operations.