Sorry, did you say "this computer that only cost me $150 a year but allowed me to test and deploy for its platform on native hardware was a waste of money"? Because it kinda sounds like the completely opposite?

As a counter anecdote, I've done this three times for macs so far, and all of those still work for their intended purpose. The 2011 model even survived overheating due to user error and fried its internal graphics cables but works fine with an external monitor. And sure, they're all stuck on whatever their respective latest operating systems are, but that's fine: they're test/compile machines, not daily drivers. And not all of them: in 2022 the oldest realistic mac target you're looking at is the 2015 macbook pro (i.e. "the good one", that even today's models can still take design lessons from).

And here's the best part: if something in them breaks, replacing the entire machine is only a few hundred bucks. Which you don't have to: you just need to buy a "for parts" with the bits you need still working. Any model that's still being used in the wild still has a decent sized for parts market. You can tell when the world stops using a model when the parts dry up.