Turns out their eMMCs are defective. Here's what's happening:
The eMMCs ignore the first read instruction, and then work from the second one on. On Linux boot, initrd makes a read request that never returns, and hangs.
If you boot off of a mSD, it woris well. You can still mount the eMMC, but you cant boot with them.
I did receiver a report that someone found a bootable (doesn't ignore first read) eMMC. But that was 1.
(Disclosure, I opened a PayPal dispute and sent back the eMMCs of my own expense. But the never responded so I got all $293 back. This goes along with 0 communication even when I was trying to do the right thing. I'm also trying to start a hardware business, and chose this platform because of no embedded radios, etc, and 1gigE on its own SPI bus.)
Tldr. Defective hardware, don't buy.
Do you have any more information or a link where to follow up on this? Skimming the threads in the forum
has more then one person saying that they managed to boot from the EMMC.
edit: nevermind found it
edit2: Are you sure you didnt run into a problem with Armbian? Did you test it with any other OS? There seem to be more people who had an issue which sounds similar, not being able to boot from EMMC but from SDCard with armbian.
Yes, I tried eMMC boot on every emmc capable OS that the Pine-fork of Etcher suggested. And I also tried openBSD as well on suggestion from freenode #pine64 user, to no avail.
I then was able to order a usb3->eMMC adapter, and was able to debug the USB and determine my assertions were correct: those eMMC chips ignored the first read operation.
I didn't have a jtag for the allwinner chips, but I also did have a serial console. I'd like to say I went above and beyond for testing what I'm defining as defective merchandise.
Now, the A64-LTS, the rtc batery holder, and the POE injecters are pretty darn good. I was just hoping for better overall performance with using eMMC.