Add section on circuit breaking behavior

Author: michaelasp

keps/sig-api-machinery/5647-stale-controller-handling/README.md

@@ -244,6 +244,20 @@ occurred. This will have the same exponential backoff semantics so after a few
 reconciles of being unable to catch up the requeue will take longer and longer
 until the cache has enough time to actually catch up to the writes.
 
+### Circuit Breaking Pattern in Controllers
+
+With the same pattern, we can implement a circuit breaking approach to certain
+controllers, such as the node-lifecycle controller. There are scenarios, such
+as the obtaining of node lease objects, where controllers may pull stale
+information from the cache, causing the controller to wrongly believe that a
+lease is expired.
+
+We can perform live gets on these resources to ensure that we don't overreact
+to staleness in the cache. Simultaneously if it is determined that the cache
+is stale, we can use the same pattern where we mark the cache as not ready
+until the cache has at least caught up to our prior live get. This will prevent
+too many requests hitting the api server.
+
 ### Test Plan
 
 <!--