As the title indicates, this question is regarding a scenario in which a coordinator needs to be replaced on an already established network. Is it possible to clone a coordinator on a new chip and resume network operation completely uninterrupted (i.e. without the need to recommission end devices)? Assume the trust center resides in this coordinator, this network has standard security enabled (NWK & TCLK), and there are bindings and reports configured on the joined end devices.
I've read numerous posts on various development forums indicating this is theoretically possible by saving network parameters in NVRAM and reloading them onto the new coordinator. However, it doesn't seem like there is consistent success (if at all) in getting this done.
Is this possible? If so, what are the high level steps involved using the ZBOSS stack, and what caveats/gotchas/limitations/ should one be mindful of when implementing this at a production scale? My understanding is, a new coordinator can restart the same network configured on the old coordinator with the same PANID, NWK, and TCLKs & short IDs of joined devices. With this information, the coordinator should be able to send packets to end devices (e.g. turning on/off a switch). Where I'm confused is how end device to coordinator communication is restored. For example, you physically toggle a switch, and it's configured to report that state change back to the coordinator. The application and NW layers will package the frames correctly to address the coordinator with the correct PAN ID, address (0x0000), NWK, etc. But what happens at the MAC layer? The new coordinator will have a new MAC address. How will the end device know of this change so that it addresses the coordinator correctly at the MAC layer? Assume this end device is a 3rd party device, for which we have no control over its firmware.