I cannot speak for Paolo, but I can speak of my experience as a GCC contributor. In the case of GCC, the FSF would do to the GCC community a great service if the copyright assignment process was streamlined as to make it less obnoxious. The current procedure is either too much a hassle or even simply impossible to fulfill. Read the horror stories of potential contributors chasing down who can sign the CA form in their organizations in the thread "Why not contribute to GCC" in the archives. I have my own horror story in which after two failed attempts and several mornings wasted, *I* proposed to the FSF the modification to my CA form that got accepted to be signed by my university.
The other glaring example is the issue with generating documentation from code. In GCC, there is a lot of stuff that is auto-generated from description files. It would be ideal to generate the documentation of this stuff from the same description files that generate the code. But the incompatibility between GPL (code) and GFDL (docs) means that it cannot be done. GCC devs have proposed many many potential solutions for this issue, but they have been either shot-down or ignored by RMS.
The remark made by Paolo about the coding standards and C++ is not that every GNU software should be re-written in C++, but that since some GNU software is (like GCC), the coding standards should acknowledge this. However, the strategy of RMS is: "I don't like it, so I am not going to do anything about it and hope you get discouraged and switch back to C." This is the same strategy as for the two issues above: "Not a problem for RMS, so RMS won't do anything about it."
Finally, there are the issues with the lack (or direct obstruction) of technical leadership and the lack of communication, misunderstandings, or failures of diplomacy and tact. There are plenty of examples of this given in the email by Paolo.