Many cassettes are built of multiple sprockets and spacers. At the risk of moving away from the original design objective for this motor (easy to fit with no mechanical changes to the bike), maybe there is a place for a drive ring that replaces the innermost sprocket with a single assembly. This might be accepted by users who would re-stack their existing sprockets to achieve a wider gear range anyway.
I think the clamping can be improved so that fiddling with the cassette is not necessary. You need a few tools to remove and remount a cassette that we might think of as normal to have, but even-more-normal-than-us people might not: cassette locknut wrench insert, cassette locking chain whip/pliers, or even a torque wrench.
How do you find the torque sensor?
Seems to be working fine for me. But maybe a simple throttle would as well. As I said before, I have only a little bit of experience with harsh uphill starts on a borrowed e-MTB. I have not experienced that with the BikeOn, but then I haven't tried starts at full motor power. Otherwise I really have nothing to compare to.
I did a quick test today to check whether I solved the spoke interference problem (sadly, I did not), but I definitely can say now that the motor output torque does increase with human torque: the interference is due to a secondary (or tertiary, depending on how you count) moment resulting from the motor torque, and sadly it scales (gets worse) with my pedal pressure. (By secondary I mean a moment that occurs due to a constraint placed on the primary moment, in this case the motor wanting to climb upward.)
Aram should comment if he wants to, but I think he is making good progress with maximizing clearances and also adding a stiffer lateral motor restraint. Turnaround will be a few weeks, as is usual for machined parts.
I think it would help if somebody with a modern 1x12 drive train could test the system as well. (Of course one could argue that Aram should be that person...

)