[en] This paper summarizes a recently completed study that identified and investigated the difficulties and limitations of applying first-order adjoint sensitivity methods to reactor accident codes. The work extends earlier adjoint sensitivity formulations and applications to consider problem/model discontinuities in a general fashion, provide for response (R) formulations required by reactor safety applications, and provide a scheme for accurately handling extremely time-sensitive reactor accident responses. The scheme involves partitioning (dividing) the model into submodels (with spearate defining equations and initial conditions) at the location of discontinuity. Successful partitioning moves the problem dependence on the discontinuity location from the whole model system equations to the initial conditions of the second submodel