Yes I think it would generally be regarded as best practice to do any data summation (i.e. stacking) before any data chopping, as it is the summation stage that clarifies what is data and what is noise, and therefore doing it the other way round risks throwing away data.
In planetary imaging in twilight or daylight (with filters) we often chop off the sky background level to increase overall contrast, but I do this after sharpening.
But another (odd-sounding) technique which I think does have some legitimacy here is to attempt to guide the way Austostakkert! stacks the images by using as a reference frame an image that has been manipulated, by masking, to be the shape the planet is expected to be. I have seen another other observer get good results on Mercury by this method.