I believe the EBDG is the type of programmer that uses two ways comms, which would indeed make it impossible to connect to multiple chips at the same time.
However, the schematic for the Zero (I haven't checked this on the Zero Pro) shows that the 10-pin Cortex Debug header - not the one to program the EBDG, the other one - is connected as it should be to PROGRAM the on board D21... Which also means it's in parallel with the pins the EDBG uses to program it in the first place.
This means it's (probably) possible to program the on board D21 with this header, but I also remember seeing some 0R resistors somewhere along the line. I believe they are connecting the lines from the EDBG to the D21, and removing them would take the on board D21 out of the equation, whilst keeping the 10-pin header connected to the EDBG.
If this is the case, simply removing these resistors would allow you to target an external board (as I believe was the original intention). It's then down to the EBDG firmware as to whether it can target only D21G18's, the entire D21 range, the entire SAMD range, or even a wider range than that.
The Zero Pro schematic is different from the Zero, however if it was designed properly then it should still have the 0R resistors - otherwise the second 10-pin header would be completely redundant.