As with all of our processing, we do things by the gigapixel of imagery. If you have 100 images that are 20 megapixels each that is 2 gigapixels.
Downscaling things doesn't really mess things up, it just provides an output that has a larger GSD. A lot of times people are processing things at way more detail than is actually needed. This gives the option of spending less to process the same data set without having to resize the imagery on your own and risk messing up the tags.
If you downscale things by .5 that same 100 images would be 1 gigapixel.
So with a Pro subscription that lets you process up to 3 gigpixels of imagery without any point charges you can either process 150 20 megapixel images with no downscaling, 300 with .5 downscaling or 600 with .25 downscaling.