Thanks for the schematic. It looks like it is properly set up. You don't need the 1 Ω resistor from the MOSFET sources to the GNDA/B pins, but it shouldn't be the source of the issue. The output channels of the ADuM3223 are two separate silicon die, so the latch-up issue sometimes seen in HV-level shifter gate drivers isn't a concern here.
I recreated your setup as shown below:
Since I'm not doing fast transient testing, the leaded parts are OK. It's hard to tell in this picture, but the drains of the MOSFETs go below the other leads, and don't short. Here, channel A is being measured with a multimeter, and channel B is being measured with a scope.
Here is the setup from further back, with the high side held low,and the low side held high. The green line is the lowside VIB, and is attached to the 5V VDD1 node. The yellow line is the highside VIA, attached to GND1:
I was able to get 6V to read out on the multimeter, but I had to put a switching signal on the input. Here is a picture with the lowside held low, and the high side VIA being fed with a 50% duty cycle 100 kHz square wave. This is occurring because the bandwidth of the multimeter is being exceeded, and it is averaging the 12 V square wave on the output:
Is it possible the 6 V output comes from a switching waveform in your setup?
Can you detach the VIA and VIB connections from the Arduino, and force them high and low manually? I know you mentioned that you slowed the time scale down to 10 sec. have you verified this with the multimeter on VIA to GND1?
RSchnell