Does anybody know how frequently one could expect Route 53's latency-based routing data to change for a given user?
For example, if I have Route 53 LBR setup with CloudFront sitting in front of 2 separate regions of my application, and R53 sends my user to the West Coast, how long is that good for? Another way of putting it is: how sticky is that decision?
Do we run the risk of the user being switched a few minutes later to the East Coast even though West Coast is still running, merely because of some minor routing change on the public internet?
Or is the decision more abstract, and rather based on IP ranges rather than directly on the user's actual point to point latency between them and the region's border? Meaning that the user is sent to the region that "should" have the least latency, but that minor real-time changes in latency will be ignored?