more about power. in this case power is torque*speed. Torque sensors (aka torque load cells) almost always use strain gauges, which measures the stretching of a structural part. This translates into a force, which can be used to calculate torque. There are also piezo sensors that would work, but that's a lot more money. Back to torque* speed. When my trainer locked up on me, the speed went to a very low level. Since I was in a mode where it wanted to control power, that meant the resistance went way up. So there is a trade off between torque and speed. If they have limitations on max torque, and min and max speed, that leads directly to a limitation on power.

Once you pick a range, then there are other things that determine accuracy. But with a strain gauge system, if you pick a huge range, as a practical matter you are left with fewer volts per watt to measure with. After all, there are very few cyclists that aren't weight weenies to some degree or another.

I think that people react strongly to this question because it's presented as a joke. Plus, very, very few cyclists ever put out a kilowatt, much less 2. Strava would know the approximate percentage of such monsters. When I was highly trained, many moons ago, I tried to put out a horsepower. In those long-gone days, we didn't have power meters, so I ran up some steps. I think I did manage to get fairly close to a horsepower, which is about 3/4 KW. Pretty sure I could put out a kW on a bike, but I've never bothered to get a power meter