It requires a loop in filters or the framework,
that makes the scheduling less efficient and more complex.
This is purely an internal change since the loop is now
present in buffersink.
Note that no filter except buffersink did rely on the requirement.