When I worked at the Open University many years ago I remember hearing of a statistics textbook that was published with numerous errors, throwing students into panic as sums did not work. When the second edition was published errors continued to be found, and also in the third. However, the canny boffins turned the situation to their advantage by showing that the diminishing number of errors still found in each edition followed a neat curve. They challenged students to use that data to calculate how many errors might be predicted to be still present in edition four, five and six.
I was reminded of this by a recent post on Paul Luna's blog which includes a nice non-apology for numerous errors by a seventeenth century printer (the point being that the book took so long to print that facts that had been true when a particular page was printed, were no longer true when the book actually went on sale).