It should be noted at a rather general level that parallel forking is replete with race conditions, and is a nasty and unwieldy aspect of SIP.
The pain of using parallel forking to implement something like "ring groups", for example, is often not worth the savings in architectural complexity.
If you have some other means of accomplishing what you want to accomplish, e.g. using a B2BUA to generate multiple independent call legs simultaneously ringing multiple destinations, I strongly advise you to go with that instead.
-- Alex