In short: not as much as you might think. All moons with water ice shells over liquid water oceans, and even those with no liquid water oceans but warmer and thus less rigid ice beneath a cold ice crust, modify craters after their initial formation via isostatic rebound.
Imagine what would happen if you suddenly excavated a crater-shaped hole in a liquid water ocean: the surrounding water would rush into the void and fill it up, until the water surface in the former void came to the same gravitational potential as the surrounding water. Ice does the same thing, only much, much slower, and this is isostatic rebound. Unlike the liquid water the really cold ice does have some strength, along with, maybe, a decrease in the porosity of the newly-formed ice in and around the crater, making it denser. That combination can stop the rebound just short of returning to a flat surface, and you get mild impact-generated depressions like Valhalla.
Research on crater formation at Callisto has been going on sporadically for decades. Recently Bjonnes, Johnson, & Silber found that at Ganymede, as crater diameter increases, at ~20-30 km diameter the depth bottoms out at ~1 km. Larger craters are no deeper. Slumping immediately after formation, and isostatic rebound later, smooth things to the ~1-km level. Callisto's cratering behavior will be similar.
If the crater's interior is completely isostatically compensated, i.e. it behaves as if it and everything around it are floating on whatever is beneath, then the depressed surface elevation will be echoed by a raised base at the bottom of the ice layer. It's similar to the iceberg phenomenon: if a cylindrical iceberg extends 100 m above the sea surface, then it will extend about 700 m below the surface.If it extends only 50 m above the sea surface, then it extends only 350 m below the surface. For Earth-like pressures, that ~1:7 ratio holds (US Coast Guard's number, for sea water; in pure water it's more like 1:9). Water and ice densities change a bit with the huge pressures encountered on these thick-water-layer moons, but if there is full isostatic compensation, a 1-km depression on the surface might mirror a 5-10 km rise in the ice shell bottom below it.
That's still small change compared to the ice shell thickness, which has been estimated at anywhere from 100 km to 150 km, though at planetary science meetings I've heard interior modelers say it could be even thicker.