If the route is [whatever us transit] -> china telecom -> [whatever connectivity in south korea], the US transit provider is generally going to send the traffic to china telecom on the US side of the undersea cable, then it can use china telecom's provisioned bit of bandwidth.

Here's the quote from the article:

> One such example the authors give is the diverting of routes between Canada and South Korean government sites. For six months beginning in February, 2016, internet traffic was diverted by China Telecom and routed through its PoP in Toronto, then forwarded to its PoP on the West Coast, then on to China and finally to South Korea. The shortest route for this traffic would normally have been Toronto to the United States to South Korea, the authors says.

But there's no cable from US to South Korea listed on the submarine cable map https://www.submarinecablemap.com/

And they didn't show any work -- it's not unreasonable to me to go through China instead of Japan, especially since BGP is generally picking routes based on the least number of ASNs in between, rather than shortest distance or anything. Probably China Telecom had a peering or transit arrangement with the destination network (or one of their transit providers), and the other available routes would have had to switch networks one extra time -- or maybe it was even, and the China Telecom session was up longer.