Fwiw, I’m not a particularly large or heavy rider.
Zwift has rider height and weight, so probably enough information to make a pretty good estimate of speed. Would not expect it to be exact, but getting a bit closer to realism, plus or minus the trainer setting, seems like a reasonable objective.
I think it will only work for either small or large gradients not both though. There are too many other factors at play other than just increasing/decreasing the gradient for it to be accurate for all gradients.
I could be wrong though so I’ll leave it with you to figure it out.
Ha! Well, fortunately, others have already figured out the math. How it’s implemented in Zwift is another matter.
But, if trainer difficulty went up far enough, that would probably cover it!
Update! This appears to be related to the resistance control sensor I pair to. The ANT+ sensor shows the strange symptoms as described. When I pair the control through the phone, however, I get substantially different (more) resistance when going up hill.
This suggests to me that there is something wrong that either Zwift or the KICKR bike are doing with resistance in the ANT+ connection that is handled correctly in the phone pairing/connection.
Some sort of bug, who knows where, but at this point using the phone pairing is a fine workaround, so there we go.
As an aside, I notice that if power is not paired through the phone, there is no option available to pair resistance control through the phone. That seems ok, but I have not seen a similar restriction between ANT+ and Bluetooth pairing, where mix and match seems possible. Not necessarily a problem either way, but curious.
And for what it might be worth, I do not see a Bluetooth pairing option for the KICKR bike for resistance control at all, though I do see one for power reading and cadence, in addition to ANT+ and phone pairing for all devices.
In any case, problem largely solved, but it suggests that something is working not quite right related to the ANT+ resistance control in my setup, if not perhaps in others.
Cheers!
In the real world you are fighting gravity. In Zwift magnetism. Or better, electromagnetism. Both fundamental forces. The total amount of energy required to climb a certain percentage/slope will be correct. However, there are differences.
True! However, if the experience is different through different sensors on the same setup, and there is apparently no documentation describing this difference, perhaps we’re looking at a bug or something like it?
indeed, you could call this a bug.
I totally agree with you that trainer difficulty doesn’t reach far enough. If you don’t shift gears and maintain cadence, your speed should remain constant (on a surface without slip) regardless of the gradient climbing up. But there are so many variables in the real world that need to be mimicked with limited (affordable) equipment. Guess to keep the experience fluent (and pleasant), there are choices to be made. In the end, it’s a simulation. But a useful one. I used (and compared, with power meter pedals) Zwift to determine which gears to mount for the big climbs in France (the only ones available at the time). Worked fine. You could compare it with astronauts. Training under water is different from working in the vacuum of space. But useful 
I didn’t read the rest of your comment before reading this out of context. 