The last decade has seen several aspect language proposals for UML 2 sequence diagrams. Aspects allow the modeler to define crosscutting concerns of sequence diagrams and to have these woven with the sequence diagrams of a so-called base model, in order to create a woven model. In a real-world scenario, there may be multiple aspects applicable to the same base model. This raises the need to analyse the set of aspects to identify possible aspect interactions (dependencies and conflicts) between applications of aspects. We call a set of aspects terminating if they may not be applied infinitely many times for any given base model. Furthermore, we call a set of terminating aspects confluent, if they, for any given base model, always yield the same final result regardless of the order in which they are applied. Since confluence must hold for any base model, this is a much stronger result than many of the current approaches that have addressed detection of aspect interactions limited to a specific base model. Our aspects are specified using standard sequence diagrams with some extensions. In this paper, we present a confluence theory specialized for our highly expressive aspect language. For the most expressive aspects, we prove that confluence is undecidable. For another class of aspects with considerable expressiveness, we prescribe an algorithm to check confluence. This algorithm is based on what we call an extended critical pair analysis. These results are useful both for modelers and researchers working with sequence diagram aspects and for researchers wanting to establish a confluence theory for other aspect-oriented modelling or model transformation approaches.