I feel your frustration, it's something I have experienced many times. It is not uncommon to turn on an old machine and find that it works, only to then find it dies a few minutes/hours/days later. All hardware is slowly decaying, even when switched off. Heat and oxygen slowly act to breakdown insulation, dry out capacitors, oxidise contacts etc. What often happens is that the heat and stress of being powered up after many years is just enough to push some components over the edge.
It is of course possible that it was the process of swapping components that did something but I think it's unlikely if you were careful. After all, it was working immediatly afterwards and it only stopped after 30 minutes. This suggests that something else has gone over the edge.
Now you have clarified that you have a 320008 board, that tells me that you have 6550 RAMS. The original test of swapping I&J4 with I&J5 and getting more memory tells us that one of the chips in column 5 is duff. If it still worked I would have swapped them individually back again or swap them with the display RAM to find the exact culprit. It's critical to have good RAMS in the leftmost sockets as they hold the zero page and stack. Having the othes missing I think would at least allow the machone to boot. To answer your earlier question, yes the ROMS and RAMS do run quite hot.
I assume a dead ram would give a blank screen, and a working ram would give at least the screenfull of PETSCII characters.
No. A blank screen is one full of spaces (0x20) so it needs good RAM. If the rams are present, but uninitialised you get garbage, as it renders whatever values the static rams have (maybe someone else here knows what a 6550 typically has on initialisation?) There is many causes of a garbage screen of which RAM, ROM and the CPU itself are just three.
Did you re-check the voltage rails from the regulators? Check all 4, the voltages should be no less than 4.9V and no more than 5.1V. If these check out good, I would reseat all the chips again. Just heating up after many years could have made them shift in their sockets a fraction.
Rob