Hi Tony, thanks. I can make a small program that will trigger this issue this weekend, I had to cannibalize the card for the rpi 3 because the project must go on, but new cards are arriving this weekend.
I will say that I discovered more, though. It's a gpxx potato, apparently there are a few different kinds. I also managed to track the fault across kernel boundaries and right into kernel space. The EINVAL is coming from the GPIO_V2_GET_LINE_IOCTL
ioctl, and while I don't understand exactly how it works, the problem *seems* to be that the GPIO22 (chip 0, offset 10, header pin 15) line is also the TEST_N line, and for some reason this line is in a different bank than the kernel pinctrl code is expecting.
I think this is true *only* for this particular line, and while it won't get in the way of most people, the adafruit hat uses this line, so it's kind of impossible to make that work properly at the moment (I actually made some really good progress at getting the rpi-rgb-led-matrix software to work using gpiod, but this is not surmountable at the present time.)
If this is not enough info, check back Sat. or Sun., I will post a small proof of concept program that will just exercise that line.
Thanks for getting back.