In this paper, we consider wireless mobile sensor networks under extreme environments where nodes: 1) have local knowledge; 2) have limited computational power; 3) make distributed decisions; and 4) move rapidly over time. Information dissemination in these networks (or gossip) can be modeled via epidemic models that analyze behavior of the system mimicking the way diseases spread (or even gossip for that matter). However, the limitation on computational power and energy of nodes forces us to consider explicit stopping criteria that are seldom done in the literature. Furthermore, harsh environments considered in this paper prevent nodes from transmitting sensed information at specified time slots and hence might cause a large variation in intertransmission time distribution. The objective of this paper is to characterize the dynamics of the information spread and obtain performance measures based on stochastic modeling. We start with modeling information flow using a Markov chain and obtain performance measures such as time to transfer information and fraction of nodes receiving information. Then, we provide a method to obtain those performance measures when the assumption on intertransmission time distribution is relaxed, e.g., time-varying transmission rates and nonexponential intertransmission time distributions, which makes our model more realistic. We make a curious finding in that, for our proposed model, the average fraction of nodes receiving information is a parameter-free constant. We also show that our model is scalable and effective.