I used to work in an environment that required high reliability. Hot-swap spares were the norm and that includes software. If you were updating software, it was done on a second (or sometimes even third) device to ensure that you always had 1 working production device and often a working spare as well. This philosophy is required in some industries, but many have moved away from it due to the cost. For a long time, downtime was forbidden. Now consultants or MBAs model the cost of downtime and can often justify it as critical savings in the short term. My local hospital recently rolled out a software update that brought down all their OR suites for over a week due to a driver issue.
I've been using Debian (and offshoots) for over 20 years and *nix before that. Original sysadmins were taught the importance of backups before updates as the OS itself doesn't provide a mechanism to roll back. There is an opportunity to provide better instructions to users on how to migrate a system safely, but it is a lot of work and there is always some risk.
I have linuxcnc running in a
NixOS
VM for simulations and might move my main controller to this one day.