and even if one has to do such a refactoring what is the point of using YAML at all ?
It allows the CI engine to determine which jobs to start, what their steps are etc.
To be fair I have worked on one project that had very complex CI and we almost decided to generate the CI graph procedurally with a Python script (Gitlab supports this, somewhat awkwardly). But in the end we decided it wouldn't be worth the overhead of writing, maintaining and learning a whole new CI system on top of Gitlab's CI.