Nope. We're most likely get dm-vs-md situation all over again.
I.e. Tejun Heo will go on and commit his "insane mode" fixes, tailored specifically for systemd. Namespace people would look at them and run away in horror - there'll be no simple delegation mechanism for cgroups anymore.
Then Google would look at this mess and decide to stick with the old multiple hierarchies interface.
So we'll get the worst of both worlds - two semi-functional interfaces co-existing at the same time, with almost the same functionality. Oh, and the single hierarchy won't make these pathological corner cases go away.
>Similarly to that delegation has problems because a delegated entity, like a container, dosen't have enough visibility into the whole system to set proper priorities as the priorities are evaluated relative to one another and are systemwide, only the administrator of the root context has enough knowledge to pick values here that can't cause interference with other workloads.
It's EASY right now to work around the weight manipulations affecting sibling groups simply by creating additional tree level. Problem solved.
And there'll STILL be required a mechanism for subtree delegation, even if systemd is used. And this mechanism would also need ACLs, security policies and discoverability. I.e. it would need to duplicate the existing FS functionality.