Why? I agree if you have chosen an ill-fitted technology stack where those aforementioned developers who didn't care about quality before LLMs tainted the training data then you do have to carefully read the code to watch for the LLM going crazy, because they will, but if you have chosen a suitable technology stack where the training data is overwhelmingly of high quality, then what need is there to read it? You already know that the quality will be on par with what you would have written yourself.
You might find it to be an interesting read, but the software will be fine if you don't.
You might find it to be an interesting read, but the software will be fine if you don't.