The "trouble" with such a warning system is that at its base is he presumption of positively identifying the infringer. Sure, and ISP can say "such-and-such activity violates our ToS, and you're responsible for all traffic on your account", but identifying a particular individual's traffic (from a content-owner's perspective) seems to be difficult given the increasing frequency of IP-as-identification being tossed from U.S. courts.

So, the biggest practical question in my mind is how they plan to reconcile penalties for individual infringers against the rights of the account holder.