When a workflow has several paths in it, there is no mechanism to control the execution of different paths. All paths are started in the order determined implicitly by the workflow engine. There are two issues with it.
1) It would be nice to be able to control which branches are run, so that the same workflow can be used to demonstrate different aspects of the experiment. This is a particular problem for the Charon analyzer because only one copy of it can be active and if multiple branches use the analyzer for different experiments, they have to be executed sequentially. To some extent, this feature can be provided by a fake "pause" analyzer. But...
2) More importantly, execution of multi-branch workflows in non-deterministic. Not only the order in which branches are started is decided randomly, but also progress of a branch may depend on the progress in another branch. Thus the "pause" analyzer can deadlock the whole workflow if the engine decides that it wants to continue with the paused branch.