Here we define and model the notion of k-resilient distribution of decisions, implemented as computations, over dynamic and constrained multiagent systems (like in IoT systems), and devise a self-organizing algorithm to repair the distribution as to ensure the system still provides the required services and remains resilient to upcoming changes. Resilience is based on the concept of replicates of computations, so that decisions hosted by disappearing nodes can activate on other nodes. The contributions are evaluated on scenarios coming from the distributed constraint optimization domain (DCOP), where computations are decision variables or constraints which are distributed over a set of agents. We experimentally evaluate the performances of our repair method on IoT-structured multi-agent systems which must keep solving a problem using the MaxSum algorithm while agents are disappearing.