Pedestrian it makes sense for your whole system to be managed with a language - language is the best tool management of complexity.
In theory, not in practice. A language is complex, you want to hide away complexity with more complexity, it's exactly what's wrong with technology today, from huge compilers to garbage collectors to the current state of web development (and even app development) with framework after framework after framework.
Matter of fact is Nix doesn't "fix" anything, mainly because most people don't think there's anything to solve, and because Nix isn't simple enough of a solution, most people aren't going to want to learn a programming language just to be able to configure their systems, and companies that would benefit from it are still just going to pay other dedicated companies to do it for them just like they already do with SQL databases.
This is likely the reason the only people you see using Nix are ricers, i.e. people who are constantly reconfiguring their systems and who need the "bleeding edge" and atomicity because they're constantly breaking things. The industry and regular users favour unchanging stability, hence why distros like Debian and RHEL are so popular. Atomicity in those kinds of environments isn't actually worth much, anywhere else, they're already using docker.
Nix also can't fix Linux binaries, because there is no solution to the problem. The varying package formats and build systems are a result of old standards and corporate interests. It's never going to be resolved unless RedHat fully takes control of the ecosystem (something they're already pretty close to doing truthfully) or some other similar circumstance which manages to mostly centralize Linux. You can see the same story in Wayland, systemd, upstart, mir, flatpak, snap, etc. Corporations making their own in-house programs and avoiding as much as possible other solutions. You may see this in Wayland vs X11 and systemd vs the world especially clear. RedHat could have rewritten X from scratch almost twice in the time it has taken Wayland to begin being adopted, but why would they? It's their in-house display server that they're making for their in-house desktop for their in-house operating system, the Wayland model benefits them more than the Client-Server model of X, they have no reason to not abandon the rest of Unix.
It is not. Complexity vs Simplicity is not related to elegance inherently.
YaST itself may be complex but it is at the end of the day a tool to be used by the user, and in that front it's a glorified settings manager and installer. Gentoo's portage is what I would rather compare Nix to when speaking about simplicity in software, but its territory isn't as extensive as Nix'. In package management I still believe it wins however, it's written in Python so anyone can contribute to it, the configuration itself is intuitive and resembles other systems people are likely to be familiar with, binaries aren't a problem because you compile everything, and compiling isn't a problem because of the USE flags. It's both elegant and simple.