Yes, I'm aware of the basics involved here. It's more that I'm not 100% familiar enough to say that the sensor actually measures strain from torque applied, or some sort of analogue that can be assumed to be akin to torque. I'm assuming they are using something like a rotary torque sensor which should be fine, but when cost is involved, I hesitate to trust my assumptions.
But the point I made still stands. If the torque sensor is a small portion of the algorithm, then the small variation that exists based on the crank arm length matters even less. It's not got a lot of inputs into the system it can even use, and so far all the ones I'm thinking of don't really care about the crank arm length. Torque is about the only one where crank arm matters at all, but in the bike use case, it's not a big factor.