It's true that 3d printing isn't going to solve all your problems; but some are likely to prove hairier than others:
Microchips are a pretty nasty case. Between long development cycles and the demand for mil/aero rated and otherwise hardened versions, military gear is quite likely to be riddled with already-obsolete parts by the time it is formally declared 'finished', much less when the Block N variant is still in use 30 years later. Unfortunately, fully accurate emulation of even relatively feeble digital ICs can be fairly tricky(just look at how much effort it takes to get a 100% binary compatible emulation of the NES' less-than-heroic 1.8MHz 6502; never mind newer stuff or analog/mixed signal); and even painfully obsolete IC fab processes are orders of magnitude smaller than alternate fabrication technologies are good for.
Boards and wiring harnesses are also less likely to be amenable to 'just press print and away you go'; but unless you've destroyed the schematics and for some reason can't tell your multimeter minions to trace it out; such relatively large assemblies should be easier to reverse engineer if necessary and rebuild as well or better with modern parts.
Parts, depending on their size, may or may not be amenable to direct 3d printing: if you go with the really fancy processes, smallish parts with comparatively obnoxious-to-machine properties might actually be easier to print than to produce by the original methods. In other cases, you might not 3d print the parts directly; but you could use 3d printing to greatly speed up the re-creation of tooling necessary to fabricate parts(sintered copper, say, is not terribly useful as an aerospace material; but if you need some tooling in stainless steel or another material that's a pain in the ass to machine precisely, being able to sinter copper to your preference, and then do sinker EDM could save you a great deal of time.)
I suspect that some older designs, unless we consider them worth a fully reverse-engineering, are now too ill-documented to be revived; but given that any current design(and probably some moderately old ones) do have CAD representations produced during design and construction, suitably robust printing technology, in combination with some other techniques, we aren't going to just nanofab them all in one piece), does hold promise.
Except for ICs, not sure what is to be done about those. Given economies of scale in the IC market; it might actually be easiest just to adopt the brute-force-and-ignorance approach and order 100x or even 1000x as many as you need, when constructing a system approved for service, just in case it ends up lasting a long time. Yeah, it will look wasteful, and some of the time it will be a waste; but economies of scale will soften the blow a bit; and you'll be saved having to source second or third hand ICs scavenged out of e-waste by the Chinese and fraudulently re-marked as new old stock; which is probably a good thing for system reliability.