quite a few OSes have the concept of 'background time', actually. in unix it's called
job control, usually used in combination with task priority control (AKA
nice).
if you're referring to the process scheduler in every modern preemptive OS - that's not what was meant in the article (and of course, iphoneOS has a robust scheduler, like every modern preemptive OS). the task scheduler does not decide the inherent priority of the tasks - it just tries to make sure the already-established priorities are adhered to. other, meta-level scheduling agents decide what task should run at what priority level. such an agent could be the user himself, e.g. 'both my mp3 encoder and my dvd ripper run normally at the same priority, but right now i want the mp3 encoder to hurry up so i'll boost the encoder/demote the ripper' - this is not something a task scheduler could decide.
AAMOF, meta- (or macro-) level scheduling is one of the fundamental Achilles heels of today's preemptive OSes - the kernel's scheduler has no/insufficient knowledge of any meta-level priority dynamics. that's why we try to predispose the kernel scheduler toward one behavior or another, based on meta-scheduling concerns. eg. in linux there are different kernel-build-time settings that allow you to chose among a server-tuned scheduler, vs a desktop-tuned scheduler, etc - these are all things that only the human user can decide.
in this regard, an 'activity-oriented' platform can be a paradigm shift in the problem of task scheduling - something along the lines of implicit job control (i.e. the job control is done by the system, based on the fact that the user interest lies in known processes). a single-task platform is an example of such a 'activity-oriented' environment - the job control always knows what the user cares about.