The repeated execution of workflow logic is usually modeled with loop constructs in the workflow model. But there are cases where it is not known at design time that a subset of activities has to be rerun during workflow execution. For instance in e-Science, scientists might have to spontaneously repeat a part of an experiment modeled and executed as workflow in order to gain meaningful results. In general, a manually triggered ad hoc rerun enables users reacting to unforeseen problems and thus improves workflow robustness. It allows natural scientists steering the convergence of scientific results, business analysts controlling their analyses results, and it facilitates an explorative workflow development as required in scientific workflows. In this paper, two operations are formalized for a manually enforced repeated enactment of activities, the iteration and the re-execution. The focus thereby lies on an arbitrary, user-selected activity as a starting point of the rerun. Important topics discussed in this context are handling of data, rerun of activities in activity sequences as well as in parallel and alternative branches, implications on the communication with partners/services and the application of the concept to workflow languages with hierarchically nested activities. Since the operations are defined on a meta-model level, they can be implemented for different workflow languages and engines.