ETAOIN SHRDLU was the standard ordering on linotype keyboards from the 1880s, so it seems unlikely the corpus it was based on was analyzed in the 1960s.
Norvig’s article doesn’t actually attribute ETAOIN SHRDLU to Mayzner’s study, which was more interested in developing statistical models for bigrams, n-grams, and letter positions. In fact it seems more likely to me that if Mayzner’s 20000 word corpus managed to match the ETAOIN SHRDLU sequence it might be because they actually used it to calibrate the corpus to ensure it was representative so that they had some faith in the n-gram analysis.
Norvig’s article doesn’t actually attribute ETAOIN SHRDLU to Mayzner’s study, which was more interested in developing statistical models for bigrams, n-grams, and letter positions. In fact it seems more likely to me that if Mayzner’s 20000 word corpus managed to match the ETAOIN SHRDLU sequence it might be because they actually used it to calibrate the corpus to ensure it was representative so that they had some faith in the n-gram analysis.