After a lengthy bring-up debug session, I noticed the #SSD1303 not quite behaving as I expected.
Pin D2 (SDA_out, which "should be left open" in #SPI mode) was not staying high all the time, like I had measured previously.
Rather it was shifting out some kind of gibberish data.
This was not compatible with my tested idea of "just putting a diode between SDA_out and SDA_in (D1)", to still allow I2C-lows through in #I2C mode, but block the constant* high signal in SPI mode.
* formerly believed as
So for rev. 1.0 I'll just remove the diode when I want to use the display via SPI or close the solder jumper when I want to use it via I2C ^^"
But apart from that, everything worked out fine. The level shifters (even up to 32MHz* SPI clock), boost converter, pinout, artwork and mechanical dimensions ^w^
For rev 1.1 (if I get around to it), I'll probably put a FET in series that gets switched by that solder jumper, but I still need to test that approach.
*64 unstable probably due to breadboard