A big part of the problem is the massive (almost exponential) increase in the complexity of the equipment over just the last few decades. And software. So much

A big part of the problem is the massive (almost exponential) increase in the complexity of the equipment over just the last few decades. And software. So much that used to be achieved with discreet parts has moved into microcontrollers, FPGAs, etc. which even if the customer had the bitfiles/firmware for, often after loading that (which might require special equipment), it then often requires configuration and calibration etc. that is very specialised (and might even need external equipment...). The amount of training required to get to that point is just massive. It would take us six months of having field techs putting units together and working with the more experienced guys to work through problems with systems in the factory before they could manage going out to the field to do basic tasks. But even then, the systems were so software-based that eight out of ten times they'd need support from the software team to diagnose and fix the problems... As a systems architect (this was in SATCOM, small specialist vendor, mostly selling to Navies), I tried to push in the direction of designing the systems to have a small number of interoperable LRUs (line-replaceable units) that the customer could keep spares of and change themselves, then ship back to us for refurbishment, as well as redundancy so the system could keep working when one of a type of unit had failed (for when it was mission-critical). I think it's a pretty good middle ground. But the problem of course is vendor lock-in where the manufacturer screws them by making replacements extremely expensive. There's not enough recourse against the big companies doing that (like banning them from other acquisition programs). Too much lobbying, back-scratching, revolving doors between military and industry goes on for that I guess...