I quite agree with everything you've said in this and your other post.
My development environment? : My own dual-booting Windows/Linux PC with 32G RAM and a few TB of SSD. Not to mention the Nvidia RTX graphics card for gaming...
I either spin up a VM to test stuff, or spin up a Python virtualenv. Postgresql also running on this machine. Whatever's needed. Need to emulate Stuff Happening From Different Servers? Why just spin up another few VM's - assign them the minimum resources required to get them doing what they need to do, set up your VM network etc. Any decently specced desktop machine can do that, never mind a noisy rack system - considering they're way better and vastly more powerful then the PC's we had 2-5 years before that, which themselves were vastly more powerful than the ones before them, and so on...
Result? Can develop at home to my heart's content, then when it comes to deployment spin up a remote VM on e.g. DigitalOcean and take it from there.
At the end of the day, "sErVeRlEsS" (I just don't like that term, for some reason it rubs me up the wrong way, perhaps because of...) just means "running stuff on someone else's kit" - the same as "tHe ClOuD", so if I'm going to be developing some system & software, I'd rather be doing it locally, setting up whatever's needed to get it running, and once satisfied, deploying it.
Like you, I see either the same people, or new people, simply Not Learning From The Past. There are many good reasons why things were done like they were - developing on a system you own, for example, rather than spinning up all sorts of Cloudy Things or "serverlessy things" right from the start.
Hardware is cheap - you don't need a supercomputer to run the beginnings of your latest Supah Scalable System[tm], you just develop and run it on a reasonably up to date box, and, sure, when you get to the stage where you need more space/bandwidth/whatever, that's the point where you deploy to some Cloudy Thing or SeRvErLeSs Thing.