I think that what I was seeing was probably just garbage from the USB OTG adapter that I was using. That would explain why I couldn't get the baud rate right.
I built a cable that should have worked the same way, without using the adapter, and got nothing back from the device.
MTK devices usually don't use the ID resistor method on their debug cables, but I did try that as well with various resistances - without success.
If I could figure out what multiplexer these devices use, I might be able to get somewhere, but as it stands, I think it is a dead end. If they left UART open on the device, it could lead to a root solution - and it doesn't appear that this has been investigated.
The boards on both the 2020 and the 2022 have pads marked RXD and TXD, I tried connecting directly to these without any success, so they must be turned off. I attempted to turn UART on via fastboot, but any OEM commands I tried were locked (not surprising)
If anyone has a USB C breakout like what is used for Google debugging; it would be interesting to see if it did anything. I don't have one, and don't really need one.
If anyone has any other ideas, I am willing to use my devices for testing. The 2020 model is pretty beat up by my kids, and it won't break my heart if I kill it. I got a pretty rocking deal on the 2022 model, I'd rather not brick it, but am willing to take some risk.
I couldn't accept my failure - so I decided I had to sacrifice the 2020. I tore into it - and spent an unreasonable amount of time getting to know it with my multimeter. I stuck my probe where few have probed before!
I found an area around the main chip with an interesting bit of shielding and though to myself - you sneaky old amazon...what are you hiding here?
I found a nice little row of 4 pins...which have a hardwire connection to the USB-C port - on pin #3, or #6 depending on which direction you count from - and ground.
I am now 100% certain that the UART to USBC connection exists. I don't own a cable that outputs that pin - and unpowered my USBC-USBA OTG adapter does not output that pin - I assume that when it is powered with the resistors there must be enough crossover somewhere to see that a signal exists, it is just too corrupted to understand by the time it hits my serial/USB adapter.
I did throw my multimeter on the visible TXD testpoint on another 2020 device, and it did show some rapidly shifting voltage up to 1.8 volts during boot - so I assume I must have damaged the other one when I was soldering my jumper to it, so it must be outputting something. Also, there are RX0 and TX0 on the back of the logic board - when I get my replacement testing board I will investigate those for anything interesting before I fry it with random components(see below)
I plan on purchasing a USBC cable with all of the pin wires, and hooking it up to see if we can get rx and tx - Maybe Console? I can't tell for sure, but it looks like the same row of pins are present on the 2022 model, so this probably translates to that device as well.
Also, of interest, there is a post here on XDA of a person who got their hands on an onyx development device. There is a small component present on that device, as well as the one in the FCC auth photos, that is conspicuously missing on our production devices - It appears that it was soldered on, and then removed(on the production devices). There is also a cable connector that has been removed - I couldn't find anything too interesting, other than some oscillating voltage up to 1.8V, so maybe another UART? This applies to the 2022 model as well.
As near as I can tell - the missing component that I am interested in, is a diode, although I can't identify what kind exactly. The pins don't ohm out to anywhere that I can tell, but the upper pin does draw down my multimeter - as if it is grounded, but the pin is not ground. The lower pin seems to be leaking just a little bit of power - up to around 1.8 volts before resetting.
Just thought vomiting here - but I wonder if this component is essential to accessing BROM and bypassing the efuses that prevent one from entering BROM via short (Which BTW I am pretty sure I shorted every test point on the logic board, some do nothing, some return you to preloader, and some just completely prevent powerup (like CLK). I didn't find any that caused a brick that wasn't fixed by battery disconnect.
I plan on purchasing another 2020 board to test my diode theory on. If anyone has a vulnerable 2018 HD8 and a non-vulnerable 2018 HD8 it would be interesting to see if we could find the same cluster of components on their logic boards to compare.
Part of the reason that I am so interested in the missing diode, is that this component cluster appears largely unchanged from the 2020 model to the 2022 model of the HD8, and if it does lead somewhere interesting, it would be a pretty easy hardware mod - as far as such things go.
I attached some images below with the interesting stuff circled.