I don't think there's testing that initial infections and initial recovereds don't overlap, or that they are subsets of the entire nodelist. This may include the case when rho is nonzero and initial recoverds is defined.