Programming on the Web enlights some classical problems encountered on large distributed applications with a particular emphasis on dynamic changes. In that context, we are interested in the questions of supervision and diagnosis. Our approach is based on true-concurrency models and consists in building an unfolding of a model of the supervised system, that selects the histories that explain the observed alarms. In this paper we extend the notion of unfolding of high-level Petri nets to a model of dynamic systems that we define. This model is close to high-level Petri nets and allows us to model dynamicity. Finally we explain how to use unfoldings of dynamic nets for the diagnosis application.