Parallel gateway Error event does not end the process

[lars1]

I have a parallel gateway, and in each branch there is a service task. Both can throw a bmpn error which are catched by a boundry error and result in a error event.
In the attached test case the first service throws such an error while the second executes normally.
I would expect that the process ends after the error is thrown, but as you can see in the test, the process is still active (stuck in the join parallel gateway).
Is this incorrect behaviour or should I setup my flow differently?

[jbarrez]

It's tricky, the bpmn 2.0 spec clearly states that the incoming sequence flows need to be counted. So one way to fix this is to replace the parallel join by an inclusive gateway (who checks the valid executions)

The workaround is to wrap the service tasks + gateways in a subprocess and put ONE error boundary event on that subprocess.