Not a direct answer to your question, but the very existence of LLVM as a serious competitor to GCC is evidence of RMS's leadership failure, and even if you ignore everything in this letter, the GCC Steering Committee should be looking at whether he's helping to steer in the right direction.
There was a fascinating thread here the other day about how LLVM was offered to the FSF, copyright assignment and all, and that it was rejected: https://news.ycombinator.com/item?id=26535789
LLVM differs from GCC in two major ways. First, nontechnically, it's under a weak copyleft instead of a strong one. Second, technically, it has a modular architecture. There's a libLLVM that you can use to parse programs and do all sorts of transformations on them.
RMS refused to create an equivalent libGCC and even to structure GCC in such a way where maintaining a private fork with a libGCC would be easy, because he feared that proprietary software developers would use GCC's code to handle the hard part of compilation and not free their proprietary extensions. He thought he scored an early victory on this front when he got NeXT to release their Objective-C frontend for GCC publicly and contribute it back. But it was, of all companies, Apple themselves (the effective successor of NeXT) that invested into LLVM.
It turned out that not only is developing a compiler within reach for a proprietary software company, developing a better compiler was easily within reach when freed from the artificial constraint of making it non-modular. LLVM was able to compete with GCC because it had a technically better design that GCC rejected for political reasons. And, ultimately, that technical excellence is now threatening GCC's dominance.
It also turns out that many software projects are willing to license their code freely anyway, without the requirement of strong copyleft. Rust probably would not have been anywhere near as successful if it didn't have libLLVM to build on top of and had to start from scratch. And Rust is under no license obligation from LLVM to be free software, but it is nonetheless free software. Apple continues to contribute many things (though not everything) to LLVM not because they have to, as they did with GCC, but because it's technically better for them to work with the public project.
So RMS's political constraints on GCC were ultimately ineffective for the free software movement, because the very proprietary developers he was worried about found an alternative, and they're releasing code anyway.
RMS being "uncompromising" would be a virtue if he were effective and talented in his choices of where he's uncompromising. But he's not.
(There are other examples of this: the lack of free software strategy for the web because he has an uncompromising refusal to use the web and therefore to learn anything about it; the failure of Ogg Vorbis as a format because of the strong copyleft of its reference implementation, which the GNU website now mentions as a lesson to learn from and avoid; the relative lack of uptake of the AGPL, and the users who would have wanted it like MongoDB and Elastic instead choosing a non-free license; etc.)