It's not rare among really good writers, Mencken is saying a really good writer is what is rare.
"I make much of this improvement because it is of an order not witnessed in American writers; and seldom, indeed, in those who start off with popular success."
Emphasis mine. American writers and the one-hit wonders of the literary world tend to get worse rather than better, he's saying. The first is probably a justifiable statement at the time—decent-to-good American literature was rare (though not entirely absent) in the English corpus before the moderns (many of whom preferred to live in Europe, anyway).
I haven't found that to be true, at least amongst successful authors. Certainly not enough to call it 'the usual course.' Most authors who I have seen be successful in early life don't write their best works in middle or late age.
I suspect it's some combination of that and:
- They had a reserve of good ideas and used them up faster than they could generate new ones
- Their initial efforts had a lot of time to "cook" in their heads, while the later ones were forced out (similar to what you propose, only it's their internal editor that's slacking off)
- They lose their enthusiasm for the profession, but they don't want to give up the success and identity they've built in the field
OTOH, I think that quality of work (judged, as it always is, by subjective appeal to some critic or set of critics) with constant skill has considerable variation in most artistic endeavors, and what makes a name is usually not the kind of work that most people can turn in day in and day out. Saying that skill goes down after someone becomes famous is, I think, a lot like saying people's skill at the lottery tends to go down after they hit the jackpot.
I think the quality and complexity of the ideas also matters, perhaps more even than quantity. Some of my favorite 'creators' who become better over time tend to focus on essentially the same thing over and over again (more than others, it seems). But their later works are richer and better and developments from earlier works.
Then there's the cultural impact he's had. Stranger Things was basically a Stephen King novel and even directly name-dropped him.
I would argue that among really good writers, it's not rare at all. The usual course is for an author to keep learning for years, figure out what works well for them, and perhaps lose a little ground towards the end of their lives due to failing health or lack of care.
Similarly, good technologists build a workshop full of techniques that experience has given them, do their highest quality work in middle to late middle age, and then retreat from active building to teach, critique or advise.