A rule about refereeing long papers

An $n$ -page paper (or proof) has probability $\ge 1-1/n$ of actually taking $\le \log_{10} n$ pages.