I wonder if it's shit like this connector that was the final straw that caused EVGA to nope out. It does appear that Nvidia is
forcing AIBs to use the connector, and that Nvidia make the only, flawed adapter. Coincidentally, I made a whole bunch (48, I guess) custom PCIe power cables for my mining rigs; I cannibalised about 15 bags of Corsair Type 4 cables and individually re-pinned the wires into new Mini-Fit Jr. plugs:
Realistically, I wanted peace of mind that I couldn't get from AliExpress/Amazon pre-made cables because I couldn't trust the crimp quality or the wire gauge used. On Corsair cables that come with their PSUs, I trust both - so I spent the time reconfiguring them to my needs. Having dismantled or reassembled multiple pins in almost 100 Mini-Fit Jr. plugs, I think it's fair to say that they are
adequate but not more than that. The pins, the crimp mechanism, the receptacle, they're all fine for the ~4A per pair 12V pin they're rated for. I don't know if I'd trust them with more than 8A, and I sure as hell wouldn't be happy doing what Nvidia's done with their 600W adapter which, after Igor's lab teardown of the damn thing, seems to push up to 16.7A down two of the pairs.
I'm not sure about 8A per pin on the standard Mini Fit Jr connector. Nvidia are using over 8A on every pin of the
even smaller HPWR pins. That's bad enough IMO but the fact that they've doubled-up the last two because 4 connectors doesn't integer-divide into 6 pairs neatly means that almost 17A is being asked of a pin that might handle 5, on a good day.
I think the saving grace for Nvidia is that of all the 4090s sold, the number that are regularly drawing the full 600W is small, and of that, the number drawing 600W for long periods is even smaller.