The development of the BioPAX is focused on conversion of interaction databases. There needs to be enough interest/examples/content in these databases to warrant new developments. The Pathway Commons project has converted (and continues to convert) many of the large exist interaction databases, and BioPAX 3 continues to prove sufficient. As such there is no set timeline for BioPAX 4, however, there are some features and unfinished proposals e.g. cell-cell interactions, friendlier semantic web ontology structure, extensions for layout and SBML embedding are also on the feature list, but not directly part of BioPAX. There are some solutions for the latter, but ideally these would be updated and standardized. At the moment you should go ahead and use BioPAX 3 and get in contact with BioPAX developers if you run into issues for your use case through the BioPAX Google Groups forum.