Ohhhh, I think it's the later ATX 2.xx standards not playing nice with ATX 1.xx standard stuff. Originally you see the board had to draw less than 10mA in standby mode, then pull down the power good, i.e. short it, to turn on. Now any logic that's gonna run off 10mA and pull down a pin is only gonna manage a few milliamps sink capability too. So the ATX 2.1 allowing for a beefy 2A can probably barely detect a low mA sink over the hum it's picking up off AC supply wiring. Might actually be a timing thing, hit the button at the peak of an AC cycle and it turns on, otherwise it doesn't.
So for this and later PSU compatibility in a robust manner, I guess we need to think of a way to amp up the powergood pulldown to reliably fire up later PSUs from older ATX standard boards. Possibly a capacitor could do it, provide a kick, but you'd need to charge/drain it through a resistor to limit it to something that doesn't back EMF the board logic too hard. This solution would mean you could only plug in the PSU, flip the physical on switch then have to wait a few mins to reliably fire the board as the capacitor would need to charge up off the PG line... IDK, spitballing this as I go, ppl will probably see problems with it.
Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.
