In extremely low duty-cycle sensor networks, end-to-end communications cannot afford to maintain an always-awake communication backbone. Low duty-cycle, accompanied by the unreliable nature of wireless communication, makes it essential to design a new data forwarding scheme for such networks, so as to achieve network energy efficiency, reliability, and timeliness in an integrated fashion. In this work, we introduce the concept of dynamic switchbased forwarding (DSF) that optimizes the (i) expected data delivery ratio, (ii) expected communication delay, or (iii) expected energy consumption. DSF is designed for networks with possibly unreliable communication links and predetermined node communication schedules. Interestingly, we reveal that allowing opportunistic looping can actually reduce the end-to-end delay. To our knowledge, these are the most encouraging results to date in this new research direction. In this paper, DSF is evaluated with a theoretical analysis, extensive simulation, and physical testbed consisting of 20 MicaZ motes. Results reveal the remarkable advantage of DSF in extremely low duty-cycle sensor networks in comparison to three well-known solutions (ETX , PRRxD  and DESS ). We also demonstrate our solution defaults into ETX in always-awake networks and DESS in perfect-link networks.