The reality is that GCC is stagnating because a lot of things people want to do are just not very easy to do with GCC, due to architectural decisions that were made for purely political reasons (and these arguments have been going on since long before LLVM even existed).
It used to be that academics, students, post grads and hackers who wanted to do something interesting and new with a compiler would start with GCC. That hasn't been true for some years now. Pretty much all the interesting new projects are based on LLVM. And that isn't because of Apple paying people. It's because LLVM has modular design that gives you a stable clean interface to various things while GCC only offers unstable and complicated internal data structures.
So increasingly people interested in studying and developing compilers are hacking on LLVM and GCC is left only with developers who have a very strong political agenda that motivates them to contribute to GCC. Unfortunately for RMS most compiler developers don't work for free out of their basements. They are well paid and work at large companies and so any emotional or political arguments about freedom don't get any weight with their bosses compared with the ability to actually get shit done. So they are switching to working on LLVM.
As detailed elsewhere in these comments RMS knows full well that the political decisions limit the functionality of GCC. He says he is happy to pay that price to prevent anybody using GCC in a project that isn't GPL licensed.
The technical problems with GCC’s unstable plugin API can be fixed, and is, from what I understand, not something put there intentionally.
Back in reality none of these projects are going anywhere. GNU zealots have spend the last 2 decades warning that BSD style licences would result in people making closed forks of things leading to the original open version dying. But it has hardly ever happened.
Instead of propagating paranoid conspiracies you should just accept the truth which is that BSD style licences have proven to be the most effective at getting people to contribute to open software.