REPLACE("string with another string in it", "string", "STRING", 2) ; "STRING with another string in it"

I think Glenn is right, but the example is wrong. Should be
"string with another STRING in it"

So is offset what we really want though? Seems like it might be either count(like Split and Join behavior), or skip(skip x number of results before the first replace).

TBH I'm really trying to think of a time when I would have ever used any of these when I am replacing.