We know at least one Commodore PET will be working late in the 23rd Century.
(By virtue of its incredibly heavy steel case no doubt.)
On the subject of electrical parts as "wear items", computers newer than the early 1980s will pose some interesting challenges to future collectors who actually want to run the machines. ICs do "wear out": thermal expansion/contraction, cumulative damage from power supply transients, even the literal "erosion" of the conductive layers over time. Very, very old computers (like the aforementioned Commodore PET) which are constructed almost entirely of "generic" 74xxx series ICs will probably be repairable for a long time, as most of the parts in those machines are still in production *today* and could/can be scavenged from all sorts of places. But you go just a few years newer, when all those discrete ICs get condensed into large-scale application-specific chips that were manufactured *just* for the computer maker to use in a few models, and you're essentially taking about irreplaceable components.
Yes, in theory you could replicate those sorts of chips in a FPGA with enough work (and in fact there are people using FPGAs and CPLDs to generate replacements for some no-longer-manufactured components like SID chips, although the quality of said replacements varies) but that's way, way *way* more complicated than replacing a few caps and a 7400 or two. I sort of doubt that work's going to happen to rescue "random middle-aged Macintosh II series machine XYZ", particularly when you factor in how poorly documented the hardware of those machines tends to be. It's a *lot* of work to make what's essentially a paperweight whether it works or not do its thing.
Anyway.