In that scenario you need to find someone completely familiar with the old model of computer to break out the multimeter and oscilloscope. That's gonna be a huge pain.
And in the meantime none of their stuff works, though I suppose they can rest on the existing library of punch cards they most certainly have. Or do it manually again.
For non-electronic parts: that's why you have a machine shop.
The electrolyte isn't milk. It doesn't go off.
Put it this way, I've replaced maybe a dozen "bad caps" in equipment in 35 years of repairing high-end musical kit, six of which were in one 1970s synthesizer made from very cheap parts.
But, I've done very well out of the fashion for "re-capping" equipment because I charge 300 quid before I even start to consider clearing a spot on the bench for stuff that's been "re-capped" before they bring it to me to fix.
> The only components that can't sit forever on a shelf
From there it seems to have spread out into a "recap everything" craze that is just silly.
The real fault is an 0.1μF disc ceramic, one of those little brown ones, that's gone leaky and now thinks it's a 1k resistor. Not that you'd know, with all the other iatrogenic faults the damn thing now has.
Awful. Just awful. The "it was circuitbent so I tidied it all up" thing was a Prophet 600 that I had to write off, it was so badly damaged.
Maybe, maybe not. If the machine uses the right "legacy" parts (a Z80 is a great example, but so is anything belt- or gear-driven, etc.), then it's a very good bet those parts and the people who are familiar with them will remain available for much longer than you or I will be alive.
I'm not saying that having latest or newest stuff is any better, just that using old stuff can bring some problems too.