Can anyone please explain why one needs to put async continuations on each activity within a fork in order for things to advance automatically?
Good question. I've filed JBPM-2888 to investigate this problem. I'm guessing it is related to JBPM-2787: process instances getting stuck in a fork, for jBPM3. There, setting the async attribute in the join node is part of the solution. What strikes me as odd reading the comments in this thread is that every activity following the fork needs to be made asynchronous in jBPM4.