The AFL++ docs mention some tips. These are the best resource on how to structure the CI pipeline at a high level I could find. Beyond that. You can use a local CI runner; that would mean that not only can you have the CI to periodically fuzz your codebase (following the tips above) but also use the server for the high computational power needed to fuzz for extended periods of time.

So, this would mean in return that you can and should take advantage of <insert_your_CI_here> functionality to rewrite parts of the AFL++ template. For example, some of these scripts can become jobs/stages.