How to reduce amount of object store memory on worker nodes in Ray Cluster?

Let’s centralize discussion here with the highly related question How to tell cluster launcher to not reserve so much memory for a ray worker node?