Advance Metering Infrastructure (AMI) networks are often deployed under challenging and unreliable conditions. One of the issues for the transmission of data packet in these unreliable networks is the routing of packets, because routing paths may behave differently from the time when the route is discovered to the time when a data packet is forwarded. In addition, control packets may get lost and give routers an inconsistent view of the network. While previous research has focused on designing the control-plane of routing protocols to deal with the AMI network conditions, there is comparatively a smaller amount of research on the advantages of new data forwarding mechanisms designed for unreliable networks. This paper introduces a set of data forwarding mechanisms inspired by distributed depth-first search algorithms, and designed for the challenging conditions of large-scale unreliable networks envisioned by smart-grid deployments. These forwarding mechanisms use data packets to detect loops, update routing tables, and perform rerouting of data packets through alternate paths, recovering thus, packets that would have been normally dropped due to failures at the link layer. We perform simulations based on a real field AMI deployment to evaluate the performance of the proposed mechanisms. We also provide the evaluation results for the data forwarding mechanisms that have been implemented in a real AMI network.