Many approaches exist to decide the order in which classes should be integrated during (integration) testing. Most of them, based on an analysis of class dependencies (for instance described in a UML class diagram) aim at producing a partial order indicating which classes should be tested in sequence and which ones can be tested in parallel. We argue in this article that, thanks to the specifics of such a class test order, it is possible to define an incremental strategy for testing classes that promotes reuse during testing, not only along class inheritance hierarchies.