I had a computer that had a failure of the 5V regulator, which then proceeded to put the full 16V unregulated rail into the computer for a few hours. It still worked fine, did all it was required to do but just had some odd behaviour, the self test never ran. Took the cover off, started checking power supply voltages on the test points. Start as usual from the bottom, -21V ok, -19V ok, -15V ok, -12V ok, 15V ok, then go to the 5V rail, and the meter did a range switch more than expected, I was expecting it to range from 1V to 10V, but it did 2 relay clicks. Shlumberger rack mounted meter, dating from the 1970's, and very loud relays ( but a lovely meter), and it was showing 16VDC on the 5V rail, not the expevted 5V1 or so it should be showing.
Power off, and spent a while changing all the parts in the %v regulator circuit, from the pass transistors to the zener reference, including all the resistors, and the SCR crowbar that was not exactly working. then power up, 5v back to 5V, and it still had that odd power on self test issue. Go card hunting, and eventually saw the rather obvious failed IC, a SFC5403PM3 IC, which had delidded itself. Ironically, the 30V rated open collector IC was the only thing that failed running off 16V. New card, and it was fine again.
Yes it did run the TTL off a 16V supply, cooking off the power in a massive heatsink block and the chassis, using a pair of beefy power transistors as pass elements, and this worked well.
The other computer used a 28V supply, and had 2 SMPS units in it, one providing a constant 5V supply, using some BUX20 transistors as switches, and an air core inductor as buck inductance. Each supply was identical, just the one was switched off most of the time, being enabled only at the peak of the reference phase 400Hz to power the arithmetic logic, then being turned off to both save power and keep the computer power use under 500W, important in a shoe box that had over 5000 TTL IC's in it.