The "temporary" fix that's been in production for six years

There is no such thing as a temporary fix. There is only a fix, and the date you stopped looking at it.

Every estate is held together by a few of these. The script someone wrote one night to get past a problem, that was going to be done properly next sprint. The setting changed just for now to make something work before a deadline. The manual step somebody does every morning because the automation broke and patching it kept slipping. Temporary, all of it, and all of it still running years later, load-bearing, undocumented, understood by one person who may no longer work there.

One I found was a tiny scheduled task, set up as a stopgap, that turned out to be the only thing keeping two systems in sync. It had been temporary for six years. It ran as a named person’s account. When that person left and their account was eventually disabled, as it should have been, the task silently stopped, and two systems began drifting apart with nobody aware that the glue had ever existed.

The danger of a temporary fix is not that it is bad. Often it is rather clever. The danger is that it never gets the documentation, the monitoring, or the ownership that a permanent part of your system would, because on paper it is not a permanent part of your system. It just behaves like one.

We go looking for the load-bearing temporary things, for free, read-only, and write down what is actually holding your estate up. Six-year-old stopgaps are remarkably common. So is the surprise on people’s faces when we point at one.