What happens when shit hits the fan in my exp is that I have to crack open the codebase and debug some portion of it, so I can explain it to myself in order to be able to explain what is wrong to the LLM.

Otherwise what I have found is that the LLM will add a new if statement which will handle the newly discovered issue and you start stacking them ifs. As the article mentions LLM's unlike humans aren't lazy, they will copy, paste add patches for every issue, why bother think and understand root cause :d.

So as part of our review we have a rule against that as well.

Humans will do this as well, especially inexperienced junior SWEs. Adding a new boolean parameter and some if statements here and there. After a while a seemingly simple function takes four boolean parameters that each control a little bit of what the function does.

The benefit is that a human who is a junior might need at least a few weeks to months of guidance to have a good taste of when to duplicate and when to DRY. An LLM likely already has good judgment, and your prompt merely needs to be activate this judgment.