How are source routes / many-to-one routes for end devices managed?

As you might have already observed, end devices (ZEDs) don’t receive route requests, including Many To One Route Requests (MTORRs). So how does a concentrator manage source routes for these nodes?

The answer is that parents of end devices send up the Route Records on their children’s behalf. They do this for every unicast sent up from the ZEDs, since they don’t bother to track which children have clean source routes established. Since there’s not a great way in ZigBee to know that a destination is sleepy, the concentrator just has to treat these destination like any other until the first delivery timeout (indirect transaction expiry roue error), at which point the stack will set a flag (Extended Timeout Flag) to denote a sleepy destination (which requires extra time for its duty cycle latency in responding to things.