I for one still think creating the initial versions of the GNU OS using solely the Linux kernel would be a great start. Once the system is stable enough to stand side to side with other GNU/Linux distributions, we could start the development process of migrating any Hurd specific software to the GNU OS which would be using Linux. I don't know a whole heap about the technical details of the Hurd OS, but I assume there are still plenty of userspace applications/features (any hurd specific IPC stuff) that could easily be integrated into the distribution without actually having a hurd/mach kernel present at the time (I could be wrong). As for the actual kernel transition itself, from Linux to the hurd, I don't think it would be too difficult of work to create a binary compatibility layer to help with this transition. I think it would mostly be more tedious work than actual difficult problem solving work. Of course this would all take plenty of time, but just stepping back and looking at the big picture of the entire process, I don't see any reasons not to take this path for the development of the Hurd/GNU OS.
Sorry for my terminology, I know it's probably not very accurate.
