Environmental energy is becoming a feasible alternative to traditional energy sources for ultra low-power devices such as sensor nodes. These devices can run reactive applications that adapt their control flow depending on the sensed data. In order to reduce the energy consumption of the platform and also to meet the timing constraints imposed by the application, we propose to dynamically reconfigure the system through the use of Field Programmable Gate Array (FPGA) fabric such that it executes more efficiently the tasks of the application. In this paper we present a new approach that enables the designer to efficiently explore different reconfiguration strategies for environmentally powered systems. For this we define a stochastic model of a harvesting video sensor node that captures the behavior of the node and of its environment. We use this approach to investigate the impact of different reconfiguration strategies for a video surveillance node on metrics of interest, such as the expected lifetime or downtime of the system. Then, we create a hardware implementation of an energy-aware reconfiguration manager on top of a custom multi-FPGA board. Our results show that the systems improve their processing capabilities if suitable reconfiguration strategies are defined for their respective configuration environments.