Leap second fallout

I looked for evidence of major leap second bugs and didn’t find many. The big news was AWS went down around 30 minutes after the leap second, but that turned out to be an improperly published route, not directly a leap second thing.

But DynResearch published a great graph of BGP route instability around that time, a 15x increase in BGP advertisements over normal.

Leap second BGP spike

I wonder what the cause of that was and whether it caused any problems for civilians. My guess is that it is related to dynamic load balancing algorithms assuming a day only has 86,400 seconds.

More info in this article, which also notes that 2000 networks stopped working briefly.