In theory, there is no difference between theory and practice. But, in practice, there is.
I connected a first version of PD (using GPIO
0 1, 2, 3 and 6 instead of 0 1, 2, 3 and 19) and an AD2 with ~10cm / ~4" wires on the above mentioned way.
Note: firstly I tried to use GPIO 0,1,2 and 3 but avrdude has some problem with 'sck = 0' (ie. GPIO0) ?!
There isn't any boot problem due to GPIOs.
The I2C and the serial connection between the two boards are OK
is not working at all works but it's fairly unstable. Probably those "too long" wires and the TXS0108E level shifter cause the problems.
BTW there are some parasitic logical oscillation - the pulse waveforms of some pins are very-very "dirty". A faulty oscilloscope probe caused these problems.
So it seems this arrangement isn't working too well. Unfortunately.
It was a quick "test" only. Further investigations and more spare time are needed. Is it worth it?
I'm really sorry. :-(