The rejection of R.A. Fisher’s groundbreaking paper defining variance seems to be one of the bigger mistakes of peer review:
Fisher completed his paper on Mendelism and biometry by June 1916 and submitted the paper to the Royal Society of London for publication. The referees suggested it be withdrawn. He subsequently submitted the paper to the Royal Society of Edinburgh, which with his financial assistance published it on 1 October 1918 under the title “The Correlation between Relatives on the Supposition of Mendelian Inheritance.”
– The Origins of Theoretical Population Genetics, William Provine (1971), p. 144
This paper has 2439 citations according to Google (which sounds extremely low), as well as its own Wikipedia page. I’d call that a success.
In addition to its importance in statistics, the paper was a key landmark in the synthesis of genetics, evolution, and biometry.