When I tested this behavior is the same on both Windows and Linux.

The change in preview update behavior is working as intended. In the "actual" video and in my testing, as soon as the mouse is lifted, the value updates and fit to page is unchecked; it is only while the mouse is held down and the value is changing very quickly that the preview does not update. We added a timeout to this field to avoid immediately regenerating the preview as soon as the value changes, as this saves us from doing a large amount of unnecessary work. If the user clicks and holds the button, presumably they are trying to reach some new scaling value, at which point they will release the mouse; it is unlikely their intention is to generate dozens of previews with the intermediate scaling values, and for larger documents doing this becomes very time consuming and makes the UI seem less responsive.

The fit to page checkbox not immediately unchecking is a side effect of that change. I do not think it is an issue since it always ends up un-checked eventually, and leaving it while the mouse is down means that the preview state and the state of the checkbox always match. +dpapad@ for a second opinion on this, since unlike the preview regeneration behavior it was not something we specifically intended.

I agree with Rebekah assessment overall.

Having said that, would it be feasible to manually uncheck the checkbox, even before any print preview regeneration is requested, as soon as the user interacts with the numerical input? Perhaps this would address the issue here with minimal effort.

Either way, I don't think this is a P1, so downgrading.