For one thing it hasn't failed, the free software movement has been an amazing success.
For another, RMS created the free software movement, in an era where most software was proprietary. The GPL, LGPL, and probably many subsequent free software licenses might not exist if not for him. Read his story that led him to this: he was trying to connect a printer to a new machine in the late 1970s or early 1980s, and the maker would not share their proprietary drivers, even though he was willing to port it to the new machine himself, so he reverse engineered it, wrote a new driver himself, and proclaimed his version to be free for all so others wouldn't have to face that same problem in the future, on the condition that they also share any improvements they make to it. Viewing most RMS behavior through the lens of this early formative experience makes it mostly make sense.
Most of the article is really not about the FSF but about emacs. Emacs is Stallman's own project - he started it in the 1980s and remains its mostly-benevolent dictator. He wants it to abide by his own strict principles. If others disagree, they're free to fork it, as has been done over the years. And his free software ideals are the reason they have that freedom.
I just don't see a general argument here to justify the broad title about the FSF in general.