A diff is when you compare two files -- the program (diff) creates an output that can be used to create one out of the other. In the context that I'm using it, we're comparing multiple versions of the same file (e.g. the extant version with an older backup). A reverse diff ... is still a diff, but you just sort of have to think of it backwards: at the same time that you're comparing the two versions, you're modifying the older one to reflect the newer one, but now you want as output the diff that it would take to create the older version of the file from the newer one (the reverse of the one you would have gotten, had you just used the output of diffing the two). Which actually should be easy, because it's just reversing the order of the two files.
Now I'm hungry for chips and chili con queso.