We introduce a new modelling assumption in wireless sensor networks, that of node redeployment (addition of sensor devices during the protocol evolution) and we extend the modelling assumption of heterogeneity (having sensor devices of various types). These two features further increase the highly dynamic nature of such networks and adaptation becomes a powerful technique for protocol design. Under this model, we design, implement and evaluate a power conservation scheme for efficient data propagation. Our protocol is adaptive: it locally monitors the network conditions (density, energy) and accordingly adjusts the sleep-awake schedules of the nodes towards best operation choices. Our protocol operates does not require exchange of control messages between nodes to coordinate.Implementing our protocol we combine it with two well-known data propagation protocols and evaluate the achieved performance through a detailed simulation study using our extended version of Ns2. We focus in highly dynamic scenarios with respect to network density, traffic conditions and sensor node resources. We propose a new general and parameterized metric capturing the trade-off between delivery rate, energy efficiency and latency. The simulation findings demonstrate significant gains (such as more than doubling the success rate of the well-known Directed Diffusion propagation paradigm) and good trade-offs. Furthermore, redeployment of sensors during network evolution and/or heterogeneous deployment of sensors drastically improve (when compared to equal total "power" simultaneous deployment of identical sensors at the start) the protocol performance (the success rate increases up to four times while reducing energy dissipation and, interestingly, keeping latency low).