In wireless networks, duty-cycling operations have been widely used to reduce the energy cost of RF idle listening at wireless receivers. Such operations, however, introduce delays in data forwarding because a sender has to wait for a targeted receiver to wake up. To reduce end-to-end delivery delays, researchers have proposed scheduling techniques , ,  to wake up nodes along the data forwarding path at the right moment. However, these techniques consider only one-way delivery from a sink to nodes (or vice versa), failing to optimally support round-trip network operations such as (i) query and response, (ii) command and control and (iii) data fetching. In this work, we interestingly reveal that the optimal roundtrip delay in a low-duty-cycle network depends only on (i) duty cycle period and (ii) the number of 2-connected network components between the source and destination nodes. We prove that optimality in the round-trip delay can be achieved by establishing a simple circular path and its related cords, in which nodes are assigned wake-up slots in an ascending order in each network component, and connecting these paths into a circular pipeline. We compared our Circular Pipelining (CP) algorithm with the state-of-art solutions , and the experimental results show that without using circular forwarding, existing solutions have a round-trip delay proportional to the network diameter, while CP remains a minimal constant delay of T as long as the network is 2-connected. We also implement the Circular Pipelining (CP) algorithm in a test bed consisting of 30 MICAz nodes, achieving significant delay reduction compared to three baseline solutions in the literature.