There was some feature that, due to changes in an external service, stopped working. The trunk was fixed to work with the new version of the external service, but the release branch was not. The bug report classified it as a regression, but it wasn't a regression, as the feature had stopped working in the same manner in production. It was decided that it was too risky to copy the fix from the trunk to the release branch.

Now, this new release also included the launch of a new virtual host, where having the non-working feature was not an option, so it was decided that the feature had to be disabled for that virtual host, so an ugly hack got put in to disable that feature for one set of use cases. Then, I got called in to review the hack and was told that it needed to be extended to some other use cases. It turned out that I had added a general mechanism to do per-virtual host configuration 8 or 9 months earlier for another feature, so the ugly hack was reverted, and per-virtual host configuration was changed to disable that feature for all the use cases on that virtual host.