http://joel.editthispage.com/news/fog0000000328.html
JoelSpolsky on his WebLog describes the process of refactoring one of his products: FogBugz?. He often makes comments on ExtremeProgramming, describing which parts are valuable to him. However, the article linked above talks about the reasons why unit tests were not written (and he also breaks the MartinFowler advice DontRefactorWithoutTest?). They sum up to:
1- It's easier to 'just write the code'
2- It's easier to 'hand check the html response'
and critically
3 - Use the working HTML for the unit test.
I'm still learning a lot about the areas that WardsWiki has introduced me too, but I can't help feeling that the description of number three is perfect for a regression test!
See also RefactoringLegacyCode, UnitTestingLegacyCode