Seriously, old software wasn't necessarily better (I well remember the old Mac system bomb dialogues!), but it certainly was leaner and forced not to be so complex.
Just throw in a shitty call-home activation DRM scheme and an authorisation server with 50% uptime too.
lul. My Java project with a bunch of maven dependencies and a slightly unwieldy number of superfluous classes doesn't have anything on modern startup's tech fetishism to over-complicate operating a website "at scale" these days.
When the hell did it become normal to suck down 3mb of javascript and poke an API just to display some otherwise static text on a webpage.
Basically - when it became obvious that people wouldn't pay for content and companies had to monitize the users.
It's lame, but.. well... suggestions?
Also because paying for content is ugly, hard, and not easy online. Subscriptions are clunky and micropayments have never taken off.
The shop can easily be given 3 raspberries and a dozen flashcards when the system is first installed, for almost no money. Even if the software vendor is out of business when it breaks in 10 years, they are good to go in no time at all. Any auto mechanic can swap the raspberry inside the case in 10 minutes and be up and running again.
The most important thing though is getting rid of the PC where it isn't necessary and making the whole thing an appliance. Otherwise the PC will be the weakest link in the system. It can be replaced but it's expensive and cumbersome (updates, security, etc) if all you want is the appliance for balancing axles.
I doubt there are any difficult hardware requirements for a task like this, if it's like balancing wheels it's no harder than (slow) analog sampling. I have done it with a cheap analog io card on a 386 with DOS for regular auto shop balancing.