We definitely don’t need precise lux readings (they won’t be precise anyway, it’s probably easiest to aim for arbitrary units) but it would be really nice to at least get out something more precise than 1-10. Bins that small won’t let people adjust the finer end of the scale to match their actual lighting.
I’d suggest exposing a value that goes 1-256, do the log-scaling in software on the switch so those values are at least ballpark linear to the human eye. This will be much easier for most people to understand and work with than a 2048 count log-scale value.
When selecting sensors, aim for more resolution in the low end of the scale - the difference of a couple lux will matter a lot for some people’s “the lights are off, or very nearly off” automations, where nobody really cares at all about the difference between cloudy and direct sunlight. If the max on the sensor is 500 lux, you’ll save a lot more of the scale for the low light situations where precise control matters.
As far as use cases - my existing motion sensors offer brightness that I can use to help decide “should this light come on, and how much”, this would do similarly. It’s possible to use the various automation frameworks to say “ok, of the 12 lightbulbs nearby, which ones are on, and what does that mean about whether this one should come on?” but that’s a huge pain. It’s much easier to say “it’s dark, so we’ll turn on the lights when someone comes into the room.” The tricky thing is understanding what “it’s dark” means contextually.