Sensor driver ICs operate in different modes for still and video, so probably developer culture between different trains of cameras diverged due to that, earlier in the technology.

Cameras are also still a bag of trickster type computer, lots of ASICs talking. “Main CPU” is probably still a thing. So disturbing the path from light to FAT file entry may or may not open some can of worms.

Lastly I can’t imagine a team of Linux hackers in a large Japanese corporate writing UVC gadget driver for company platform... entertainment value and frictionless operation are very neglected/disincentivized part of engineering in the country.

Maybe the second point is the most relevant. They’re not designed to encode and export video to PC and that itself is a challenge.

Ah, super interesting and that's what I was curious about.

Yes, I imagined that there were a lot of written specs requested from the group responsible for building this app/interface...

Yes, it’s a massive difference in quality. Feature films will often incorporate video shot with e-mount APS-C cameras. Live TV is often streamed using HDMI capture decks like Blackmagics, coming from cameras not too different from consumer mirrorless cameras.

You can properly control a dSLR or mirrorless. You can determine exact aperture, field-of-view, exposure, etc. That's helpful if you're doing any sort of machine vision. Webcams have more controls than normally exposed, but you still can't e.g. swap out a lens for a different focal length.