I was just thinking about the driver today. I think I'm almost caught up on work from my road trip and the kids are settling into online school finally, so I don't need to spend so much time on family IT issues. So I think I can look at it this weekend.
Which version of the Pi have you been using? Do you have any logs I can look at?
I'm not exactly sure why Bias frames would throw a problem. The driver itself doesn't do anything different for Bias frames versus other images. However, it *may* be related to requesting that the driver deliver the Bias frame in FITs format, or Kstars/Ekos (on the Pi) automatically converting to FITs, if that's applicable to your setup. It's possible that your Bias frames are inherently more computationally intensive for FITs conversion, which could cause the driver to run out of resources and crash, depending on the hardware you use. I think the Pi4 should be able to consistently handle it, but I know the Pi3 will often crash during FITs conversion, unless the camera is set to shoot low resolution JPEG images. The JPEG resolution can only be changed in PTP mode, so perhaps you have changed the resolution in PTP mode, and that is why BIAS frames work. This is all just a hypothesis, of course.
If you're upgrading from a Pi3 to a Pi4, you'll definitely see some improvement, though I don't think 5GHz Wlan will by itself make much of a difference. The main bottlenecks for me at least are 1) the transfer from the camera to the Pi; and 2) FITs conversion, if I'm doing it on the Pi (even with a Pi4). I would recommend turning off any setting that causes the driver to convert to FITs, and then set Kstars/Ekos on your local desktop to auto-convert to FITs instead.
Your bug with the AC adapter is very weird. I don't have the exact same setup to test, unfortunately. I honestly have no idea. Logs may offer some insight.
You mention having to switch to manual when you do short exposures. I was noticing some odd behavior myself when I was trying to capture Jupiter out in the middle of nowhere last week. I think this may have been the first time I had field-tested anything that needed less than a second with the Pentax driver, and it did not go as expected. Seems like there may have been two bugs. 1) I think anything less than a second may crash bulb mode; 2) it seems like MSC mode may not be requesting the right exposure time for short exposures. I was going to investigate these more when I had the time. But I'm just wondering if you observed similar problems.