Have you read the release notes of GCC 4.6.0? Have you compared the Fortran section with the C family section? See also http://gcc.gnu.org/wiki/GFortran. I would dare to say that nowadays Fortran may be more actively maintained than C and C++. The difference is that C, C++ and Ada are mostly maintained by paid developers, whereas Fortran is mostly (only?) developed by volunteers in their free time. The only reason the Fortran FE doesn't improve faster is manpower, not politics or bickering. They are going as fast as they can with the resources they have.
The D and Pascal FEs are not in GCC mainline because their maintainers have not bothered to pursue the process to achieve integration. We may disagree with the process (copyright assignment, quality criteria, etc.) but the point is that steps are well-defined and little effort has been made to follow them. The same can be said of UPC. In fact, those projects are currently under-maintained, so who will shoulder the burden of keeping them simply build-able as the rest of GCC progresses? GCC devs have already their hands pretty full, do you really want to divert efforts from mainline GCC to a Pascal FE?
The argument "out-of-tree -> duplication, within-tree -> common library" is totally bogus. As an example, the Ada, C/C++ and Fortran FEs do not even share something as basic as the diagnostics library (there is not even such a thing, but ok, they don't share the code in any form). There is plenty of duplication already in GCC and adding new FEs will not reduce it.
Finally, the quality testing of GCC is quite comprehensive right now. There is a huge testsuite, automatic testers, runtime checks, testing done by the distros, etc. GCC does not require more people looking for bugs, it requires more people *fixing* bugs. Adding new FEs that are disabled by default won't help here. Bugs that only affect non-primary FEs will have such a lower priority that they will be effectively ignored, so their impact in the quality of GCC will be zero.