Currently, the lens seems to scale the current view, instead of the original image.
This doesn't matter this much if the scaled view is roughly equal in size to the original image or bigger. But if the original image is substantially large (for example, if the page is a two-page spread and the view is set to "Fit Width"), especially if the text then is quite small, upscaling the scaled down image leads to a text that's not really all that readable. Ideally, the lens should instead source from the original image.
Of course, that complicates the lens scaling math a bit. Instead of just applying a 2x/3x (or whatever qcomicbook does) do the current view, you need to take the current scale factor into account.