As a process developer, I want to enable jobs from a single process instance to execute concurrently so that a process can complete in the shortest possible time.

In the example given in the documentation (http://docs.camunda.org/guides/user-guide/#!/#concurrent-job-execution), there is a parallel gateway for "book hotel", "book flight" and "book concert tickets". The engine will execute these steps sequentially, even though they are modeled as parallel and asynchronous.

If book hotel takes 2 hours, book flight takes 1 hour and book concert tickets takes 30 minutes, then the whole process between the split and the join should take 2 hours (the max time of a single path) and not take 3.5 hours (the sum time of all paths).