I wouldn’t take too seriously anyone saying it’s a horrible idea. I mean, I think you could always argue it’s a waste of resources running a GUI for a thing intended to be a server. But headless servers aren’t the end all be all. I’m sure there’s a lot of licensed redhat instances out there running gnome or whatever because reasons.

Personally I wouldn’t do it unless some hard necessity were there because it’s just another thing that could go wrong, another thing to maintain if you’re capturing your config as code, and mostly because I’m not gonna dedicate a keyboard/monitor for that kind of stuff.

I’m not the op of the comment, but there is a good chance it is Semiosis by Sue Burke. It’s such a fun read, and quite unique for several reasons.

I did a 4 node Pi4 kubernetes cluster for about 5 years. The learning experience was priceless. I think most notable was learning to do proper multiarch container builds to support arm and x86_64. That being said, about half a year ago I decided to try condensing it all into two n100 nuc-like clones and keep one pi as the controller. For me and my apps and use cases there was no going back. Performance gains were substantial and in this regard I think I was hobbling myself after the educational aspect plateaued.