Patch Rollbacks

Many corporate networks treat patching with an urgency that may lead to as many problems as it solves. There are certainly some patches that are more urgent than others, but if the rush to patch means not testing the patch or doing a less thorough test of the patch which results in breaking network or application functionality it poses a problem.

I am writing an article and I am interested in hearing what sort of procedures you have in place on your network for testing patches before implementation. I would also like to know under what conditions that testing might be skipped to rush a patch into implementation and whether that has caused any problems. Lastly, I would like to hear what sort of processes or tools people are using to rollback or undo patches that are broken or cause more problems than they solve.