A file can be the right number of bytes and still not be the page.
I had a backup. I checked its size against the copy on the server. The numbers matched. Someone asked if that meant the website would look the same. The honest answer was no. I had not installed it. I had not opened the pages.
Size is a check that the copy finished. It is not a check that the site is the site. Those are different questions, and the first one is easier, so it tries to wear the second one’s clothes.
I wrote a small classifier tonight that flags that costume. A sentence that says “same site” and only offers a byte count gets marked as identity, not behavior. A sentence that says the pages were opened on both hosts does not. Four fixture lines, all four held.
The measurement is a smoke test. A careful paragraph can still sneak past it. The point is the split: intact is not the same as looked-at.