If software engineers actually burned down all the old code every three years and literally everybody had to learn this new thing in order to do anything at all, this claim would make sense. But it's all additive. Even as we here on HN are shrieking and hooting about Framework B and the way it shines a unique light on the world of programming and how no thinking person could possibly resist its charms, thousands of people keep chugging along on Framework A for years afterwards. In no small part because they've heard it before, and remember that somewhere around the 2 year mark of Framework B's run you can expect well-written and cogent articles about "Framework B Considered Harmful", "Framework B Is Permanently Ruining An Entire Generation Of Programmers", "Why I'm Switching Back To Framework A And Regret Ever Listening To Hacker News" and "Oh Shit, I Bet My Business On Framework B And All My Developers Have Turned Into Feral Wombats And They're Biting Me Ow Ow Ow".
On the frontend, React has been king for quite a long time now, with no obvious contender in sight. Of course there's stuff like Vue but the difference isn't so big that people are throwing away their gigantic React codebases to restart anew with Vue (things that did happen a lot with jquery, bootstrap and even angular based frontends). The new kids on the block may be better, but only by a relatively small margin.
Similarly on the backend, Rails has been unfashionable for quite a while now, but there's no obvious single new framework that people are flocking to. Sure, there's many hip options, like Elixir/Phoenix, vanilla Go, and bunch of "serverless" frameworks, but just as many people just start their new projects on Rails, Laravel, and ExpressJS, because that shit is productive and it just works.
I feel like, decades after the desktop UI world, the web as a platform has pretty much matured. The core concepts aren't changing that much anymore. I think that maybe something fundamentally new has to happen again for this cycle to truly restart - just like the web as a software platform was once fundamentally new. We'll get new programmers who can reinvent all these ideas again in confusingly subtly different ways. It'll be fun!
It's narrowly referring to the "framework fads" that do seem to sweep through our profession on about a three-to-five-year cycle.