use ipfs as a backend for artifacts/cache/dependencies
artifacts/cache/dependencies are keywords that gives ability to pass some files between jobs. At the current moment it looks like it is centralized in a way your downstream jobs will download those files from ci coordinator. It won't scale well for complex workflows using kind of "shared working pool". I think very good direction here could be integration of ipfs open source tech. So runner when sharing something would simply produce hash and share it.