Well, on a different note, even Linux desktop/workstation distributions have reached their Vista point... in the somewhat trivial sense that they now need Vista-capable hardware in order to be reasonably usable! (We do know how it happened, it's all about the usual CADT syndrome. At least they are still usable, unlike Vista or Catalina. But one can still be disappointed by such developments.)
Vista level hardware is 10-12 years old by now though. I don't think it's unreasonable to drop support for hardware over 10 years old for a fully-featured environment. People want new and shiny. Why should I want to give up on some eye candy on my five year old laptop just so that someone else's 10 year old laptop isn't missing out on anything?
Mind you, Firefox and LibreOffice are bigger slowdowns than the rest of my systems and I'm using vanilla Gnome3. The system can still boot a fully featured desktop with less than a gigabyte of RAM in use. Even Vista hardware should be able to cope with that.
When I bought my first EeePC I expected to get a reasonably recent gen hardware, just without the bling and low power. It had a brand new 10 year old Intel graphics chip inside. At least it had a decent battery life and doubled as a toaster if you opened a pdf in firefoxs javascript pdf viewer.
There's also the fact that many developers live in rich countries where good hardware is relatively cheap and easy to come by. Developing countries usually don't have access to reasonably priced hardware at all. There's billions of people whose hardware is barely capable of running a Vista equivalent. That alone should be reason to consider developing with older systems in mind.
Developing to keep an old system somewhat smooth can lead to improvements across the board. There's pros and cons to every decision and I certainly don't think every piece of software should be designed to run on a Core2Duo. In many cases some considerations for those who can't bug fancy hardware would be nice though.
This is especially true wrt. memory use, BTW. Memory bandwidth is, as a rule of thumb, the main bottleneck affecting performance in present-day systems, so most savings in RAM use yield very real increases in performance for real-world scenarios. Also, less RAM use makes it easier to run a useful number of virtual machines or containers, even on non-high class hardware.
The problem is, running on the CPU does not make your DE faster (it's the other way around), memory that is the largest problem isn't an issue, there is some heavy disk usage on load time, but then none, and there isn't much of a CPU load. They can't run on a 2006 machine, but it does not extend into your VM.
For example, youtube doesn't play well, because the hardware is too slow to decode a full HD video stream in real time. No kidding, that's actually a big driver for people to buy newer hardware, people really care about youtube.
Another example. Loading a modern website with 10 MB of javascript bundles. It takes 5-10 seconds to process all that javascript in chrome (on developer workstations), which is slow but fine. Unfortunately that's 15-30 seconds on laptops and mobiles for end users, that is quite a lot, enough that people leave thinking the site is broken.
An application that doesn't run smoothly on old hardware like that will probably not run well on current hardware either. It's twice as true if the application should be usable on modern low-end laptops, that can really have terrible hardware, worse than a desktop from back then.