Increasing fault-tolerance of multi-agent systems