The target model, if it's complete and ready for use on an executable platform, can now be executed. The behavior of the resulting system is the same as when the application model was tested at the end of the knowledge formalization step (assuming the mappings are specified correctly, of course).
The remaining testing work falls into two main categories. First, we have to make sure that the functions that the system performs are in compliance with the specification. Basically, these steps correspond with the normal testing routines that every software development process exhibits. Then, we must verify that the performance of the resulting system is adequate. If it isn't, it's likely that the choice of implementing abstractions, or the allocation implied by the marks, is at fault. Fixing these doesn't require any change to the application models, only to the implementing abstractions or annotations.