When and how did "fail" become a noun?

Online Etymology Dictionary says:

Fail:

The noun (e.g. without fail) is from late 13c., from O.Fr. faile "deficiency," from falir. The Anglo-French form of the verb, failer, came to be used as a noun, hence failure.

Some hypotheses are listed and briefly analyzed in this Slate article.

adding to Mr Mehper's answer, people use it in the stock and security market too. to mean:

1. Failure to deliver securities to a purchaser within a specified time.
2. Failure to receive the proceeds of a transaction, as in the sale of stock or securities, by a specified date.

The recent uprise in incidences of non-verb fail is due to its being an internet meme, possibly from a Japanese video game with poor translation to English. Know your meme.

Although "fail" has been used sparingly in the past as a noun (e.g. "without fail"), there is evidence that the word is creeping toward ordinary usage as a noun to replace the word "failure." I just read the following headline in Bloomberg Businessweek (6/6/11-6/12/11 issue): "How Stephen Elop is trying to lead Nokia past its epic fail." My first reaction was that perhaps the word is only used to draw attention in headlines, but that it would not breach the body of the article. Well, the word did appear once inside the piece.

"Fail" (as an ordinary noun) is coming into our campsite, ladies & gents. Hey, compared to "refudiate," it's a gem!

Fail has been used as a noun with a quite different meaning by Rev John Davidson,DD in his book 'Old Aberdeenshire Ministers and their people' 1895. He describes a parish school built in a day with dry-stone walls topped with FAIL or divot and a roof of open rafters covered by thatch.

The correct answer is.... The exact date when the explosion of FAILs noun usage began probably cant be pin pointed. But where and why is absolutely and unquestionably....... The online game WORLD OF WARCRAFT. And more specifically it applied to RAIDs in WOW. In a raid you gather 15 or 30 online players at a specific location in the game. You all then enter a special area where your group must conquer a series of Bosses to compete the raid. They were a pivotal part of the game. Before a raid would start the most knowledgeable players in the RAID would assess the 15 or 30 man partys likelyhood of completing the entire raid. Many fators were considered. After assessing the party's strength the leaders of the raid would declare whether or not "This RAID party is a FAIL" Meaning it is ununlikley the party is strong enough to complete the entire raid. Exactly when or who started using FAIL in this fashion in WOW we'll never know. But i started playing WOW shortly after its release. I remember the first time I saw FAIL used in this fashion. I'm thinking it was right before; or shortly after, the release of the first expansion pack. I also remember how it went from being used periodically to becoming a staple word used in the game vernacular. From there it spread like Corona-virus to the rest of the planet. The release date of the first expansion pack for World of Warcraft is Nov 13, 2008. At the time it was the #1 online PC game with about 100 million players. I've read about the Japanese Arcade game meme and do remember it. But I dont remember a focus on FAIL. the entire poor translation was the joke. And its popularity was short lived. It was the forcing of FAIL into the massive WOW community's vocabulary that gave it the traction it needed.

My answer focuses on the emergence not of "fail" as a noun but of the phrase "a fail" in the sense of "a failure."

The earliest instance I've found in a Google Books search is from testimony given by Stian Finne—"a clerk with a stock sales house"—in May 1927 in Duncan v. Stoneham, a New York state court case, where the term appears repeatedly as a short form for "a failure to deliver" or "a failure to perform:

Q. Now, you testified that Stoneham gave Clarke & Company a check on a fail to deliver or fail to receive, and you took that check. Was that a wrong transaction? A. No.

...

Q. If the delivery is in odd lots, is to be in odd lots, that might give rise to a fail, might it not? A. Yes.

...

Q. If you are to deliver stock to another broker and the broker from whom you are to receive that stock has not delivered to you yet, that would give rise to a fail, would it not? That is a cause for a fail to deliver, isn't it? A. I don't understand your mind there.

...

Q. If certain securities are out in loan of that character, and they have not come in and in the meantime you are transferring an account to which they belong that could give rise to a fail, could it not? A. Yes, that might give rise to a fail.

So it appears that in 1920s stock brokerage slang in New York, "a fail" was an understood short form of "a failure to deliver or a failure to receive." This form of expression seems to have persisted in market talk, as we see from this example in New York Institute of Finance, Introduction to Brokerage Operations Department Procedures (1979) [combined snippets]:

Under the trade-for-trade and balance-order settlement, securities have to be delivered and paid for to satisfy the firm[']s obligations. If the firm is unable to deliver securities on settlement date, the procedure is referred to as a Fail. The selling firm has a "fail to deliver," the buying firm records a "fail to receive." When the selling firm delivers the security , the buying firm pays the money due, and both firms clean up these fails.

Still, a more likely direct source of "a fail" as it is used in everyday speech today is the field of education—specifically in connection with the idea of "a failing grade." An instance of "a fail" used in this sense appears in "The Meaning of School Certification," in The [London] Journal of Education, volume 64 (May 1932) [combined snippets]:

He ["Omega"] also asserts that standardization is essential for correct results. One wonders to what extent previous consultation could be effective in such cases as those of the three candidates who are awarded, for their essay and précis, a fail and a distinction by different examiners, or of the twenty-five who are failed by one or more examiners and given a credit by others. Would "standardization" here mean splitting the difference, or would certain examiners have to abandon their convictions? And what of the candidate who accumulated a fail, a pass, a credit, three special credits, and a distinction for his essay and précis, or the two who obtained a pass, four credits, a special credit, and a distinction for their literature? Would a majority vote carry the day?

The use of "a fail" seems to have caught on in the United States and the UK across multiple fields in the late 1960s and early 1970s. Here are some examples.

From an unidentified speech in Parliamentary Debates: Standing Committees: House of Commons (1965) [combined snippets]:

The strange thing is that universities accept the B grades as sacrosanct. They will not look at a pupil who has not got more than two. But there is no check on the figures of B grades. If someone is around 49 per cent., a pass-fail, then he is examined to see if he is a fail or a pass. If he is a B grade it is not checked, in any way; its accuracy is taken for granted; possibly the difference between 59 per cent. and 61 per cent.

From John Valentine "GCE—a Comedy and Tragedy of Errors," in New Education and Programmed Learning News (February 1967) [combined snippets]:

The illusion that a grade of C on an 'A' level examination is a C, is a C, is a C when it is probably a case of a B or a C or a D, is a cruel illusion for students, and a foolish illusion for the universities that make distinctions between grades of C and B. The comfortable assurance that a pass is a pass, a fail is a fail, and a Certificate therefore rests on some bedrock of rational justice, is a very grand illusion.

From an unidentified article in School & Society, volume 96 (1968) [combined snippets]:

Still another approach to the problem will be under way at Ball State University (Ind.), Louisiana Polytechnic Institute, and California State College at Hayward, among other schools. They will offer students certain courses under a pass-fail option, through which the student receives either a "pass" or a "fail" for the course instead of a specific grade.

From an unidentified article in Milk Producer, volumes 18–20 (1970[?]):

The results of the composition and hygiene tests are entered on the laboratory record card, and a carbon copy slip of the card showing butterfat, solids-not-fat and total solids percentages and hygiene test result—"Pass" or "Fail" is sent to the producer each month. The producer is notified of the antibiotics test result only if it is a "Fail" and then on a special form MQ.243.

From an unidentified paper in Proceedings of the Institution of Electrical Engineers, volume 119 (1972) [combined snippets]:

As mentioned above, n is related to the boundary mechanisms and the form of the static characteristics. The previous analysis has assumed n = 1, but it is now instructive to examine the effect of a variation in n by choosing another typical value (n = 0) and displaying the new clear/fail curve on the same axes in Fig. 4. It is seen that the effect of the change in n is to move the the curve into the 'fail' region, and so the curve for n = 0 predicts a 'clear' for some conditions for which the curve for n = 1 predicts a 'fail'. A circuit breaker having boundary conditions giving n = O would be more successful than one having conditions giving n = 1, as would be expected as the losses in a collapsing arc will be higher in the former case.

From an unidentified paper in Society of Plastics Engineers Annual Technical Conference (1974):

The Falling Dart Impact, FDI, test is used to obtain information on the resistance to failure of polymers in a realistic simulation of end-user abuse. During this test a weighted hemispherically tipped dart is dropped onto a test piece clamped between a pair of steel rings. For each piece tested, the operator notes the weight of the dart and whether a piece ""failed" or "held". For this study a fail is considered to be a sample with an easily discernible crack. The intent of the operator is usually to determine the energy (height times weight) at which 50% of the samples fail. This energy level is called the mean break energy, MB.

Conclusions

Although "a fail" can be found in stock brokerage patois dating to the 1920s, the current widespread usage of the term most likely originated in 1960s usage of the term in the context of education—specifically, in the context of pass/fail grading. From there it seems to have passed fairly rapidly into other fields where grading is important, and then to general usage as a short form of "failure" or in the sense of someone or something that deserves to receive a failing grade.