ReferenceGlossary
Leap smearing
A timekeeping fix for the leap second: a server spreads the extra second over a long window, often a whole day, so its clock runs very slightly slow and never jumps. Google and Amazon do this on their own clocks. A jump is what hurts, because a repeated second makes timers fire twice and leases expire early.
Reliability
In Quorum#
The Leap second or clock skew Incident makes every Node in one region answer 1,000 ms late and time out 10% of its answers, until you restart the Nodes or it runs out after 60 s. Leap smearing (Series C) is the Research that keeps the jump from happening. The Leap Second Challenge scripts the same event as a regional App Server outage.