I wish I knew more about what the inversion valve actually does, but if you have air leaking from the service chamber to the QR on the relay valve when the parking brake is on, then I think it's probably a bad diaphragm in the DD3. In my schematic, the inversion valve is connected to the "lock" section of the DD3, no pressure when locked as in when the parking brake is applied, and pressure when parking brake is released. So no air is applied to the lock port to leak when parking brake is on. it's also connected to the emergency brake chamber on the DD-3. I don't see a path to the service chamber at all.
Air is applied to the emergency chamber when the parking brake is applied, and to the service chamber if you 'set' the brake with a brake application. If you don't do a service brake application when putting the parking brake on, then air is only applied to the emergency brake chamber. In either case, as soon as you release the service brake application, the QR on the relay valve operates and there should be no air pressure in the service brake chamber. There will be pressure in the emergency brake chamber, at least until it leaks off, but there is no difference to the parking brake action, once it's on it stays on.
So if you have air continuing to leak from the R8 valve when the parking brake is on, then you have a leak between the emergency brake chamber and the service chamber on one of the DD3's.
I always 'set' the parking brake with a fairly full service brake application. the Bendix data sheet recommends against this, it prefers that the parking brake be set with pressure on the auxiliary diaphragm only:
PARKING
The DD-3™ actuator parking function is activated by releasing
the air pressure from the lock port and making a brake
application. The application may well be made from the same
alternate source as used for the emergency application. It
should preferably be made on the auxiliary diaphragm for
reasons which will be explained under “releasing.”
If the air application is released or leaks off, the shaft will
retract slightly and then be clamped by the wedging action
of the rollers. The brakes may now be considered to be
applied and to remain applied regardless of any exhaustion
of energy or loss of air pressure.
RELEASE OF PARKING APPLICATION
To release a parking application, air pressure in excess of
40 psi must be applied at the lock port and a brake application
slightly greater than the locked-in application must be made.
Thus if the park application is made with the auxiliary
diaphragm, an application of equal air pressure on the larger
area service diaphragm will provide the heavier application
required to free the lock rollers and release the park
application.
Document:
http://docs.google.com/viewer?a=v&q=cache:hK_OLBFoJrwJ:www.bendixvrc.com/itemDisplay.asp%3FdocumentID%3D2393+bendix+dd3&hl=en&gl=ca&pid=bl&srcid=ADGEESjjtjTEV8aJAAI6Kx5pAuJh8KCS2JA7B-DVcI87GtHTyzPUaRmR2cS6kbq2pMbPLft2OzMpKZ7x2SebpsAi0OBB6XHsGyK5oHWHl_jDkaAXFEAS4rpmmS06LNnFnkLoutUZQbXP&sig=AHIEtbQB6PEmimHmqUwnYTpQdBypKOuVyQHope this helps, even a bit. I also hope it's right... just working from logic and data sheets!
EDIT: thinking about this more, it's probably only half the story. I do see a path from the inversion valve to the R8 valve through the shuttle valve.
Brian