 Certain disease pandemics arise when a population weakened by one illness becomes more susceptible to another. In the present study, researchers examined how the shape of a network determines whether two cooperating pathogens gradually infect nodes in the network, or spread violently through an abrupt and massive outbreak. In networks where the pathogens spread across long distances before converging on the same node, the spread of the two pathogens could wane before suddenly entering a runaway avalanche outbreak phase. The hallmark of this phase was the infection of a sizable fraction of the entire network after either infection alone would have died out. Human populations were modelled as networks of interconnected nodes with a variable number of connections per node. To start the infection process, a single node was infected by a pair of pathogens A and B. With each tick of the simulation clock, any node connected to an infected node had a chance to become infected. Both pathogens were assigned the same probability of spreading to an untouched node, and pathogen cooperation was modelled by giving a node infected by one pathogen a higher probability of contracting the other. The development of avalanche behaviour depended on how well a network could delay one pathogen from cooperating with the other. Essentially, that delay depended on how the network was constructed. Networks containing a large number of short loops allowed pathogen A and pathogen B to connect quickly. This network structure favoured pathogen cooperation and led to smooth and predictable spreading of the pathogens through the population. In contrast, networks containing a large number of long loops created long paths between A and B. This structure delayed cooperation between the pathogens, leading to abrupt and therefore unpredictable spreading. In other words, if the pathogens could manage to survive on their own until their paths met, massive secondary infections would occur because one pathogen could intrude into the infection path carved out by the other. A catastrophic outbreak was more likely to occur in networks with nodes randomly connected to a number of other nodes than in networks with all nodes connected to the same set of neighbouring nodes. This is because random networks generally contain longer average paths between pathogens. These findings reveal the potential for massive, unexpected avalanche outbreaks of cooperating pathogens in networks. The simulations explored in this study suggest that postponed cooperation of individual pathogens can suddenly give way to a massive outbreak of doubly infected nodes. This finding has potential implications not only for models of the spread of disease, but also for other interacting entities, such as financial crises. The authors point out that the cooperation examined in their study is only one of many types of agent interactions derived from game theory that could encourage global instability and therefore warrants further investigation.