Potentially, they do not have multi-AZ ELBs and only thought to have their EC2 or similar instances in multi-AZ. Or they don't have proper tuning of their DNS t

Potentially, they do not have multi-AZ ELBs and only thought to have their EC2 or similar instances in multi-AZ. Or they don't have proper tuning of their DNS to stop routing to failing ELBs. Or... Amazon is unaware or not making us aware of the side-effects that their subnet outages tend to have on the overall network in a region.