Well, that really looks like a focus shift. On a non-motorized focus/zoom cam, that can only come through a few things:
- Lens to sensor spacing changing. This can depend on how the lens focus is locked and how the lens holder is mounted to the sensor board (often as part of the IR filter). It's pretty rare for the lens to actually turn in the holder, but stuff happens.
- Lens elements shifting within the lens. I've had a few rattly lenses in my day, and a lens that worked well through QA may have had elements loosen up during shipping.
There's a bit of chromatic aberration (purple fringing) on the lens, but that's not uncommon on M12 lenses, since it costs more money to make lenses without it.
I'll be tearing mine apart and taking pics at some point, but should probably test it a bit first.
ETA: It just occurred to me that software could cause an apparent focus shift as well.
Assuming the lower resolutions show the full FOV, or something close, they are likely to sum and average adjacent pixels to get the larger pixels that would make up, say, a 720p image. If it were displaying averaged pixels in 10MP mode, this would look just like being out of focus.
I'm not sure how you'd test something like this (maybe reducing the 10MP image to a smaller resolution, though that should give the impression of better focus anyway), but if the mechanical components are solid, a software glitch makes some sense.