Scratches are mountains and valleys. You can remove all the mountain peaks and create new mountain/valleys (smaller ones for a finer stone), or just flatten the existing (coarse stone) mountain peaks, leaving the deeper (coarser stone) valleys.
It is whatever you desire
Just pick one at whatever grit you need that doesn't leave random singular deep scratches.
These two points get to the heart of the matter for me. And do so in a more pithy and eloquent way than what I'm about to say...
It is perfectly possible to get a mid grit (say mid 1000s ish) natural stone to leave a finish with no real noticeable scratches to the eye. The point of using a finer stone after
isn't to completely erase that finish, it's to build on and refine it. Hopefully keeping aspects of the previous stone's finish, a dark jigane for example, and add nuance from a finer stone maybe to bring some more shine to that, or highlight features of the steel.
There are various reasons that natural stones work better for this in comparison to synthetics. The most obvious being that synthetic stones are meant to abrade; they are designed to be more friable and have constantly exposed new abrasive, and this is particularly important when you consider what they're made from. SiC and AloX are notably harder than silica - synthetic stones inherently cut better, faster and more consistently.
By comparison naturals are harder (less friable) in composition, but the abrasive itself is softer. The slurry on a natural stone doesn't physically 'break down'; what happens is that the silica, which is only marginally harder than hardened steel, rounds out and loses its abrasive power. If the stone is very hard and non-friable you'd call it 'burnishing' the surface, and both mean that the stone becomes slower and the finish shallower as you continue to work it. In this way you're already building a kind of progression when you use a mid grit stone, and it's why even coarser natural stones like Binsui or Natsuya are good in the early stages of polishing. Or indeed for something like setting the bevels on a razor.
A further thing to note is that natural stones often also have other abrasives in them, and depending on the (Mohs) hardness, some of them might have an impact on cladding, but not on core steel. You can't simply view a mid grit stone and a finisher as coarser or finer versions of the same thing, it isn't just about removing or replacing an identical type of finish.
If a natural whetstone is very hard that is usually (not always) the result of lithification. And as well as altering the physical structure of a stone the pressure involved in lithification also compresses and smooths the silica or quartz within it - what might have been circular particles before become disc shaped. Natural stones tend to get cut parallel to the direction of their fissility, or perpendicular to that of lithification, with the result being that the cutting surface is less abrasive and the finish or 'scratch pattern' shallower. Which ties into the other reason you can't view a finishing stone as a finer version of a mid grit, because they're not necessarily. Usually they're just harder...
Very hard and fine jnats *do not necessarily have a smaller grain size*. I have a Maruo Shiro Suita, that is at least as fine as top razor stones under a scope, it's just softer and has had less lithification, so it doesn't finish as fine. The Japanese grading system according to 'hardness' rather than 'fineness' is entirely appropriate for these kinds of stones, and the result of a deep understanding of their of their composition and effect.
---
I realise that the above was slightly rambling, but all of those things impact in various ways to why I think earlier stones in a progression
do matter to the finished result. In fact I'd go so far as to say that if you use a finishing stone to completely remove any effect of a previous stone's finish then you're not doing it right.