Tech@ZWO , thank you.
I contacted the seller, and they forwarded me a response from ZWO:
=== QUOTE BEGINS
This does not affect the shooting, just use flat field stacking calibration, this phenomenon has no effect on astronomical shooting
The reason why SLR cameras cannot see it is that even in raw format, the image will be processed by built-in algorithms to reduce fixed pattern noise and unevenness, such as in-camera correction.
But astronomical photography is because the original fits image is read without any correction or processing, so this phenomenon will be seen. Flat field can be completely calibrated and will not have any effect on the output.
f the problem still exists after removing the protective glass, it is due to the unevenness of the CMOS itself, such as a problem with Sony's process. However, as mentioned above, SLR cameras will perform in-camera correction so it will not be visible, while astronomical cameras will capture the original image and this problem will be visible.
=== QUOTE ENDS
ZWO are quite right here, as the banding is completely removed by the calibration with proper flat frames. In fact, gradient banding is quite subtle: I performed image measurements that showed only 0.1-0.2% difference in brightness between a lighter and an adjacent darker band.
Why I am stuck with this thing is that I have not found anything similar on the Internet, it looks like I am the only one observing such an artifact. What I found were references to uneven brightness in the form of spots, not clear bands.,
I am also curious to find a real cause for the issue, eventhough it does not require fixing, based on the response above.
Someone on the forums mentioned that uneven brightness can be caused by a thin-film layer inside the sensor and by uneven thickness of the air gap beween that layer and other ones: this explanation looks quite realistic.