Hi Tarik,
Excellent questions. From a stability perspective, all components are open source and supported by their respective open source communities. In Open Data Hub we do have multiple components offering similar functionalities in the AI/ML phase and are driven to upgrades based on our user and community interests. Airflow in our community is not as popular as Kubeflow pipelines and probably needs an update, not being accessible from the dashboard is not an indication of maturity, it should be there, please open an issue if it is not showing up.
As for choosing which components to use, my advice to you is to start by understanding the requirements of your users, your team and the cluster available. Even though some of these components provide similar functionality, digging deeper you will find different features. For example while both Seldon and KFServing provide model serving, Seldon allows for custom model metrics reporting for prometheus and grafana while KFServing does not. On the other hand, KFServing uses Knative that includes features such as AutoScaling, Canary rollouts and batching not present in Seldon.
I think providing documentation that illustrates the different features of each component is a great idea.
Juana