Wireless environment monitoring applications with significantly relaxed quality-of-service constraints are emerging. Hence, the possibility to use rough low knowledge routing in sensor networks to reduce hardware resource and software complexity is questionable. Moreover, low lnowledge handling allows better genericity, which is of interest, for instance, for basic operation enabling system set-up. In this framework, this paper revisits stateless unicast random walk routing in wireless sensor networks. Based on random walk theory, original closed-form expressions of the delay, the power consumption and related spatial behaviors are provided according to the scale of the system. Basic properties of such a random routing are discussed. Exploiting its properties, data gathering schemes that fulfill the requirements of the application with rather good energy efficiency are then identified