This paper studies the factors that sustain mental disorders by taking a network approach. The network theory suggests that mental disorders are networks of symptoms that causally interact (Borsboom, 2017). Symptom networks share certain dynamics with other complex systems: abrupt transitions between stable states, critical slowing down and hysteresis (Cramer et al., 2016). These findings suggest that symptom networks that have transitioned to a pathological state tend to remain that state. We argue that this tendency leads to the Lindy effect in symptom networks. The Lindy effect means that the conditional probability of surviving beyond a time point, given survival until that time point, increases over time (Taleb, 2014). In other words, time benefits future survival. A symptom network is considered to have survived until a time point if it has remained in a pathological state until that point. We first show how the Lindy effect is formalised by examining the stopping time distribution of Brownian motion with an absorbing barrier (Cook, 2012; Taleb, 2018). Specifically, we describe the hazard function of the stopping time distribution and make a distinction between "strong Lindy" and "weak Lindy". Strong Lindy is a monotonically decreasing hazard function whereas weak Lindy means an inverted-U shaped hazard function. Then, major depressive disorder (MDD) networks were simulated, manipulating the level of symptom connectivity. As before, the presence of the Lindy effect in these networks were tested using hazard functions, and in addition, survival probabilities conditioned on time. Afterwards, we fit a distribution to the network lifetimes. The lifetime distribution of strongly connected networks were heavy tailed and showed the Lindy effect; the longer a network had been depressed, the more likely it was to remain depressed. The lifetime distribution of weakly connected networks were light tailed and did not show the Lindy effect. After discussing caveats and alternative explanations of the findings, we conclude that network dynamics and the resulting Lindy effect can explain several findings in psychology such as the chronicity of depression (Swaminath, 2009) and the frequency distribution of remission times (Simon, 2000; Patten et al., 2010).