In this paper, the real-time failure-recovery of flexible production systems is considered. Initially, the problem is structured and formulated within a systemic framework, and then, a generic local-in-time, schedule-repair approach is proposed. The coloured timed Petri net formalism is used to represent, in a modular and compact way, the production system's discrete-event dynamics. In connection with this, an intelligent opportunistic recovery technique has been developed for obtaining, in real time, the recovery trajectory which conforms best to a chosen production strategy. The application of the approach is demonstrated by considering recovery cases in a batch-processing chemicals plant.