I am not sure about cmd_flashupdate since it is not mentioned in the programming manual on page 17.
I went with cmd_flashwrite and for some reason it did not fail but it dod not work exactly either.

However, you could try to add a header to the SPI pins between your controller and the BT815 in order to connect
a VA800A-SPI to it and use the EVE Asset Builder.
This should work when you put your controller into reset.

This is one more reason why I stick with modules like the EVE3-50G, I can detach my controller board and connect
the module to my "EVE2 USB to SPI" from Matrix Orbital or my "Hermes Board" from Riverdi.
The current iteration of my controller board is only 45x36 mm, has a stepdown regulator, a speaker, an ATSAMC21E18A-AU, a LDO for the CAN and a TJA1051 CAN-FD transceiver.
I use a 5cm FFC cable.

According to page 17 of the Programming Guide cmd_flashfast is not needed to write to the FLASH.
A couple of commands like cmd_flasherase and cmd_flashwrite work in "Basic" state as well and this way the Asset Builder is able to write to the FLASH thru the BT81x even if the FLASH has been erased or is completely new.
My best guess is that "Basic" is using simple SPI and "Full" is using QSPI.

Well, at least in theory it is that way and I experienced cmd_flashwrite not working myself recently.
My code was doing something wrong and I still need to find out what.

The init in my library starts a little different:

EVE_pdn_set();
DELAY_MS(6); /* minimum time for power-down is 5ms */
EVE_pdn_clear();
DELAY_MS(21); /* minimum time to allow from rising PD_N to first access is 20ms */

#if defined (EVE_HAS_CRYSTAL)
EVE_cmdWrite(EVE_CLKEXT,0); /* setup FT8xx for external clock */
#else
EVE_cmdWrite(EVE_CLKINT,0); /* setup FT8xx for internal clock */
#endif

#if defined (BT81X_ENABLE)
EVE_cmdWrite(EVE_CLKSEL,0x46); /* set clock to 72 MHz */
#endif

EVE_cmdWrite(EVE_ACTIVE,0); /* start FT8xx */

chipid = EVE_memRead8(REG_ID); /* Read ID register */
while(chipid != 0x7C) /* if chipid is not 0x7c, continue to read it until it is, FT81x may need a moment for it's power on selftest */
{
chipid = EVE_memRead8(REG_ID);
DELAY_MS(1);
timeout++;
if(timeout > 400)
{
return 0;
}
}

This does not only work faster than a fixed delay, it also works with no display attached.

Today I received two SST26VF064BA-101I/SM from Microchip which I bought from Mouser.
Instead of the industrie-standard 50...180 seconds for a chip-erase these do a chip-erase in 50ms max.
I soldered one onto my RVT43ULBNWC00 and tested it with the EVE Asset Builder.
The whole "Erase" action takes less than five seconds now which really is great for testing things.

Edit: after some tries and no help with this here I gave up on the SST26VF064BA-101I/SM now.
Yes, the chip-erase is very fast.
But as it turns out, EVE refuses to write to these.
The SST26VF064BA-101I/SM are detected correctly with the right size, but cmd_flashwrite does not write
anything to the chip.
And there is no output from the EVE Asset builder to indicate what could be wrong.

I removed the SST26VF064BA-101I/SM again and put back the 64Mbit flash that the RVT43ULBNWC00 was deliverded with.
And after a (painfully long) erasing of the chip with the EVE Asset Builder my test aplication used cmd_flashwrite just fine
for a small file containing the unified.blob and two images.

I know that Riverdi has EVE3 modules, I could get them from TME.
I have a RVT43ULBNWC00 I am playing with.

However, I already have a finished design for a tablet-like case to fit an EVE2-70G in and Riverdi uses
a completely different PCB.
The need to take additional steps when converting images is far less of a inconveniance than to
redesign our case and maybe even our PCB.

Of course you are right and if DigiKey and Mouser keep pushing the delivery data I just might get modules from Riverdi regardless.

I didn't initialize the display list to any value before writing REG_PCLK. Now I send DISPLAY() and write REG_DLSWAP before REG_PCLK.

I think the behavior between FT813 and BT815 is different, though. I have never experienced any problem with my previous PCB.

Well I have a FT800 Programmers Guide.pdf from 2014 here and the given example for the initialization example is pretty much the same in the current BRT_AN_033_BT81X_Series_Programming_Guide.pdf.
And you are still not following it correctly.

I converted the notosans-regular-webfont.ttf I generated to sizes from 13 to 30 now.
I used the EVE Asset Builder 0.4.2 with its default settings and only changed the font-size from one conversion to the next.
ASTC block footprints: auto
Compression Speed: thorough
Addr of Font Data: FLASH 4096

I put all the .glyph and .xfont files in a flash image and placed an image at the end.
My programm first copies the .xfont files to G-RAM using cmd_flashread.
In the display-list my programm used cmd_setfont2 to assign bitmap handles to the custom fonts.
And finally my programm prints out the same text ten times with different fonts.
The first 9 lines are font-sizes 13 to 21, the last line is with internal font 28.
The image is displayed directly from FLASH.

I was prepared to add my files as attachments but I could not attach them all: Restrictions: 4 per post, maximum total size 192KB, maximum individual size 128KB

The high-res version of Font_Test_4.jpg is 1,09MB and the zipped Font-Test_4.bin is 219kB.

The notosans-regular-webfont.ttf is the version of the notosans-regular.ttf from Google that I stripped down to only 222 Glyphs.
This is purely attached for testing purposes and in no way meant as redistribution for anything else but testing.