Blame the Tier-1 & Tier-2 backbone providers and telcos for skimping on SONET implementations; UPSRs (Unidirectional, Path-Switched Rings) do not have the line-fault switching capabilities that a BLSR (Bi-directional, Line-Switched Ring) because of the single-direction design of a UPSR. Since UPSR networks are cheaper (1/2 the fiber-lay costs) than BLSR, many large telcos and backbone providers play fast and loose with fiber capacity and provisioning...which, in this case, apparently came back to bite them.
The original ARPANet, as it was designed at that time in history, *was* redundant and met the needs for the spec. The ARPANet / NSFNet is as distant from today's Internet as a Blue Whale is from granite.
During "The Great Internet Build-out" of the late 90's, outages similar to this were more common than what you have been led to believe; the reason why people heard virtually nothing about those outages was because (a such outages weren't "visible" to those outside of the telco industry, and there wasn't such a demand 10 years ago for such high capacity circuits, and (b circuits were more carefully planned-out and used BLSR as much as possible. Now, where stockholders go crazy if their investment in a given telco doesn't grow by 10%, those telcos scrimp and cut corners wherever they can - including running SONET networks with inherently unsafe ring topologies.
For more about the differences in SONET topologies, please visit: