Core scheduling lands in 5.14

Usually, looking out from the kernel, a thread is a process is a thread, just with varying degrees of memory-map sharing. Thread groups as processes is largely a user-space notion; and those processes collected into cgroups is another. It is easy to see why the kernel prefers to avoid the issue, and try to treat them all as an undifferentiated pile of threads, wherever it can get away with that.

Scheduling, though, is a place where it often can't, because users want what they think of as fairness.