Assuming no slowdown from all the context-switching (a generous assumption), and equal sized features, splitting the team would make the high priority feature take twice as long to deliver.