Revisiting the "Request to reopen “Literal Meaning of 'Gosh'", I observe that a few minutes of dedicated research for that question ("Literal Meaning of 'Gosh'") antedated the earliest OED attestation (1757) by 36 years (to 1721). Incidentally (perhaps), I observe the context of the earlier use of 'gosh' is Scottish dialect:
Bawsy. O gosh! O gosh! Then, Jouk, ha'e at her,
If that be a', 'tis nae great matter.
Accordingly, I have voted to reopen the question, which, although it shows no research whatsoever, attracted three stellar answers in short order, as well as posing questions for which there can be no [easily discovered] general reference answers. I quote the questions from the question:
- Is this just a corruption of the word 'God' or does it have some other provenance?
- How long has it been in use?
Question 1 is not answered by OED's unsupported etymology, which does not even go so far as to cross-reference 'gosse' (although the latter does cross-reference 'gosh'); Etymology Online, at least, mentions a "probable" connection with 'gosse', as pointed out in Callithumpian's answer, but does nothing to evidence the claimed probability (unless you count mention of 'losh' as evidence, which I do not). In short, OED's etymology begs the question wholesale, and Etymology Online's puts a retail price on it.
Question 2 reminds me, once again, that closing, as general reference, etymology questions that ask "how long", puts the cart before the horse; OED's dating (and so also Etymology Online's) is itself frequently out of date. Data available from contemporary corpora routinely antedate their findings. Often, as well, the earlier evidence suggests other origins and specific histories than are proposed by OED, especially in the case of phrases, but sometimes also in the case of single words. This is to say, a claim that OED, posited as a general reference, provides the final say on the earliest date of use for a word or phrase presumes that research beyond general reference has been undertaken and yielded no contradictory evidence. Research beyond general reference being required, before the cry of "general reference!", belies the claim that the question can be answered with general reference tools.
About the previous point, I would like to accordingly edit the question to state baldly that OED and Etymology Online do not, on the face of it, provide adequate answers. I hesitate to so edit the question, however, because the edit would render one of the answers, an honest answer to the original question, half-foolish.