It seems like each device needs to be able to talk to the hub directly at least once for this to work properly, which is sort of counter-intuitive. Does anyone have a detailed explanation of exactly how the repair works?
This has been discussed four or five times in detail in The various threads on zwave repairs, what exactly were you confused about?
The hub has a list of all known devices. During a Z wave repair it attempts to contact each device and ask it to update its neighbor table. It’s a call and response format command. (It uses the normal mesh system, so the messages do get passed along in both directions using repeaters.) If it doesn’t get a response, then it notes the error. I’m not sure why that would be “counterintuitive.“
The following thread goes into excruciating detail if you need more than that. 
The counterintuitive part would be the assertion that it has to talk directly to the hub, ie not via mesh.
Spawned by a conversation where someone was having issues with devices and the z wave repair wasn’t doing anything even after many tries. His solution was to physically move the hub so it would be able to connect directly to the devices with issues. That worked and even after he moved the hub back still did. Lending the thought that the devices have to connect directly to the hub at least once.
So…a curiosity, I suppose… 
Thanks, as always, for the info JD.
Ah, ok. No, the devices do not have to be within one hop range of the hub. They use their normal mesh network. 
As far as the person moving the hub around, without a network map it’s not possible to know exactly what happened, but probably one of the repeaters was bad and moving the hub closer told the device about a different repeater and it started using that.
3 Likes