Typically wireless networks coverage is planned with static redundancy to compensate temporal variations in the environment. As a result, the service still is delivered but the network coverage could have entered a critical state, meaning that further changes in the environment may lead to service failure. Service failures have to be explicitly notified by the applications. Therefore, in this paper we propose a methodology for fault-tolerant coverage planning. The idea is detecting the critical state and removing it by on-line system reconfiguration, and restoration of the original static redundancy. Even in case of a failure the system automatically generates a new configuration to restore the service, leading to shorter repair times. We describe how this approach can be applied to wireless mesh networks, often used in industrial applications like manufacturing, automation and logistics. The evaluation results show that the underlying model used for error detection and system recovery is accurate enough to correctly identify the system state.