Having a single tree would mean that if the x86 maintainer would update a port, it would also be synced to all the other architectures without any testing. Also, it would not be possible to have separate versions of the same port, which can be needed if version X doesn't build/work on some arch.
In addition, some ports are platform specific (e.g. lilo (x86), silo (sparc) etc.), and it wouldn't make sense to sync them to the user.
Another big drawback of any "single tree" approach is that you need commit access to that tree. A separated approach means that anyone can start with a new architecture, without being dependent on others.
Finally, I don't really buy the "easier to maintain" argument. The hard part is to get a port working. That's an initial effort. When you then do an update of that port, most time is spent on building and testing it, not bumping the version in Pkgfile. That's independent from whether there's a single tree, or separate ones.
Hope this explains it, Regards Johannes
But still "ARCH" variable could help with maintaining private repos, sometimes easy fix is needed for certain port to build, so contributing user instead of starting new repo could simply do the fix in one Pkgfile. regards Bartek