GNU software absolutely has been stagnating for the past 2 decades under Stallman's leadership. It is a little known fact that in 2005, there was an offer to merge the LLVM project into GCC, and assign the entire copyright to GNU. The patches to do so were even submitted. 
They were rejected , because GNU wasn't interested in producing a modular compiler. They thought that if the compiler was allowed to be modular, then it might be used by proprietary software behind their backs.
But it turns out that modular compilers are generally quite useful, even for huge quantities of FOSS software. So the LLVM project continues, outside of GNU, and here we are 15 years later and essentially all academic research, and all new tooling development, happens in the LLVM ecosystem. Massive self-inflicted wound.
And yet, the stonewalling  and bridge-burning   continued, long past the point of rationality. Continuing to object to development of useful tools even after the damage was done. At this point, just pissing off a lot of people who were interested in doing the work to help improve the usefulness of GCC.
And not limited to this topic, either. Such as when Stallman decided to leverage his "veto power" in opposition to removing an abortion joke from the glibc manual, against the consensus of nearly all of the glibc core team, despite not having contributed to glibc himself in years. 
And if I was interested in continuing to dig through mailing list archives, I could probably pull out a bunch more examples.
Even within active GNU projects, this egregious incidence of mismanagement is still not widely known, although many maintainers of active GNU projects have their own stories of when Stallman's ignorance of a subject led him to make poor decisions.
For at least a decade, his near-total ignorance of how people (billions of them!) use computers has been encouraged and deepened by the sycophantic acolytes who encourage him to ignore developments in computing, like the ostrich with its head in the sand.
All the while, the FSF and GNU can do less and less for the users who are stuck living in the dystopia that he predicted in the 80s and 90s.
It's as if, after the first time they booted a computer with a fully-free operating system, they declared the work to be complete, abandoning the future.
Can you expand on this point? I don't doubt you, but I'd like to know more.
The LLVM world has moved to a non-copyleft one, with results that RMS maybe considers mixed: on one hand it attracts work (but being non-copyleft is neither a necessary nor sufficient condition, and for ex Linux attracts way more work by being copyleft), on the other hand RMS does not think that helping to create proprietary software based on free software is a good idea (of course the main purpose of a compiler is to compile software, even proprietary ones, but the philosophical issues come when you can make proprietary derivatives of the compiler or parts of it, not merely use it)
In the end, for a supposedly stagnant project, GNU is not doing bad. It is still an essential part of mainstream Linux based systems. It is being actively developed. GCC is moving fast and also attracts lots of work nowadays.
OpenOffice is stagnant. GNU? Not that much.