Nearly all the Genesis scrolls are small fragments. You can’t compare a fragment to the medieval text and conclude the whole book was stable.

True of each fragment individually; the argument is cumulative. Around twenty copies covering different portions of Genesis, all agreeing with the later text where they overlap, give a broad sample of the book across two centuries and many scribes. No single scrap proves stability, but two dozen scraps from different hands and decades, none of which shows a different Genesis, is strong sampling evidence. The claim is that nothing in the sample points to another version, not that any one piece proves the whole.