You clearly and obviously have little or none of practical, genuine experience working with experienced Lispers, and driving your opinions from a book-read definitions rather from actual, battle-tested work.
Your line up of Lisp and Haskell next to one another is already telling - the two cultures can't be more different. Haskell culture historically have selected for people who enjoy theory for its own sake. Lisp culture genuinely prizes getting-shit-done attitude.
> What scales and works is simple and boring.
Java was exotic in 1996 and boring by 2006. If "boring" just means "widely adopted and familiar", then your notion just collapses to "what's popular is popular". True and absolutely pointless and empty.
I reckon, you're piggybacking on Dan McKinley's "Choose Boring Technology", but his point was that the innovation tokens are scarce. That is a good and true argument. But it's about organizational risk budgets, not language virtue. It says nothing about Haskell or Lisp being bad or impractical - it says being different is expensive, and you'd better be buying something with that expense. I personally disagree with that argument (in this case), because I have effectively proven that Lisp's bus factor is much smaller than even of Python or JS/TS, because modern Lisp dialects are far simpler and less convoluted and easier to train into. The experience difference and angle of opinions of two Python/JS experts can be dramatic. Having two Lispers talking "the same" language is far more plausible in practice. Not to mention that ROI from hiring a practicing Lisper (regardless of the business stack) more likely to exceed of a programmer without such knowledge.
Let me remind you that Python was considered a niche scripting toy at some point when it was the same age of [niche] Clojure today. It got boring by riding two waves that had nothing to do with the language design: web (Django/Flask) and then, decisively, being in the right place and time when scientific computing and ML needed a glue language.
Boring things survive by mass, inertia, and being unremarkable - COBOL endures because ripping it out of banks is too expensive, not because anyone loves it. Lisp survives by being remarkable - by smaller population who see the thing the majority can't, regenerating the flame across dialects out of conviction, not inertia.
What scales and works is boring, that is true. Lisp is not that, never will be, and doesn't care. What has enduring value survives regardless of adoption - that is Unix, SQL, lambda calculus itself. I'm sure you won't be arguing that any of these are impractical. Anyone can wholeheartedly accept every new rising and falling, boring COBOL and chase every new hype cycle, or just quietly keep feeding on truly everlasting ideas. Or, like in my case, nothing stands in your way of combining both - I use plenty of boring languages at work, and efficiently utilize Lisp for my personal computing. Because the darn thing just fucking works!