Funny, I did quite a deep dive on this same issue about two years back, and came to exactly the opposite conclusion: keep ASDF, and lean on Nix to do pinning. I now maintain my own Common Lisp package repository, testing daily against nightly SBCL etc, and I've learned that a lot of CL stability is just because test suites are never run. Nix goes a long way to shield me from the haphazard breakages that other commenters mentioned, but of course it comes with a huge downside: you need to use Nix :)
At the risk of derailing the conversation (although Guix is a lisp so maybe not): I agree 100% but also maybe Nix's pragmatism is why it's more popular? "Pragmatism" being a programming language euphemism for "untyped hacky mess".
Why is this in past tense? Guix development is active. It's just Nix is more popular and it's understandable since it's the original idea and it's older.
Guix even has a very active, high-quality blog where maintainers detail major technical accomplishments, long-term goals, etc.: https://guix.gnu.org/en/blog/
From here it seems like they're growing and advancing well. I wish I could find ready historical data on the numbers of packages and services from, say, 4 years ago vs. today, though. I could have sworn Repology used to show year over year stuff but I can't find it now.
(just kidding)
Code at https://github.com/hraban/cl-nix-lite if someone is interested.
Obviously since this was written by Fukamachi you know it will be good. Much respect to the man.