Welp… Let’s use some guesswork and deductive methods to resolve this.
So you are limited to 2k x 2k final image you want to bake texture to. Say, your object to paint to was a Cube and you have prepared T shaped UV map on previously mentioned canvas. Effectively your loosing something - area outside UV faces is a Wasteland. You are left to calculate percentage.
Now to the painting. Assume, you used 10k x 10k NASA Blue Marble image as a Stencil map and had zoomed Stencil so that you see only 4 pixels in one corner of Blue Marble. This is - you used 4 pixels from 10k x 10k image to paint area equal to one UV face in 2k x 2k image. …
Another extreme - you zoomed out Stencil so that it covers 1/10 of one face of the Cube and left black the rest. Can you calculate how many pixels there will be filled on the console screen by this splotch if this Cube was to function as a Player and would cover about 1/5th of the 4k resolution screen?
All this was to - you’re limited to 2k x 2k (you could upscale your existing 1k image(s), not that you add any more detail by doing this).
Use maximum UV area and common sense, some experimentation to see what suits you best.