Do note that Finsky being Finsky, no doubt the PCIe leads are 16ga and survived without even a blister. Then again, together 9x 16ga wires can carry a helluva lot of current. Connectors got toasted as aftermath of the main fault. What very obviously blew are the power planes on the PCB's.
He mentioned it was an s7 batch-1, those have no Vcore reg so the strings of chips are fed directly from the PSU with nothing at all to limit power going through the PCB other than the resistance of the copper itself. The first 3 batches of s9's are the same setup.
When a chip or something else shorted the power planes and with the board fed by a supply more than happy to dump in over 3x more than an entire 3-board miner normally takes
Been using the Bitmain PSU's for several years with zero problems other than the fan whines more after a couple years. While a similar fault when using a more reasonably rated supply would still toast a board the result should be far less catastrophic since the PSU would shut down a lot sooner.
Oh, and all computer PSU's and pretty much all industrial ones are switchers. Other than in legacy equipment from the 70's or for very special needs I haven't seen a high power linear supply in decades.
Had no idea Bitmain has cut down so much on costs and production quality recently (well, I did know but I wasn't aware of catastrophic failures resulting from em like this). Thank god Finsky actually found those issues in time, could have been a lot worse for all of us if Bitmain got away with it. Hope they don't try to sneak in the same thing for the next batch of their future SHA miners.
Just another note to always use good, low AWG cables. Good cabling not only prevents heat issues and melting wires but can also help diagnose the root of an issue like with the faulty S7s/9s. I personally never use any cables higher than 16s with breakout boards.
Never use any molex or SATA to 6 pin connector either, they use high awg wire and the quality on those are extremely bad.