How GitLab Shared Runner works with kubernetes

I am trying to gather deeper understanding of how the GitLab provided shared runners for k8s work. How GitLab makes decisions to assign jobs to concurrent runner when kubernetes supplied runner is used? I found this brief explanation but I find it somewhat confusing.
I was wondering if anyone has struggled through trying to understand the relationship between GitLab and the shared runner when deployed on k8s?
Thank you for taking the time to read my question.