Is it normal that my lights keep disconnecting after the network update?

The ghost entries and inconsistent responsiveness you are seeing are almost certainly related to how the devices are being identified within the Matter protocol, not necessarily a simple hardware failure. It sounds like the router update shifted how the network is assigning or broadcasting identifiers, wich is causing the hub to see multiple instances of the same fixture. This is actually an issue that the latest standards are designed to fix, but sometimes the transition phase creates these kinds of bugs. What you’re describing with the duplication is likely related to the adoption of Endpoint Unique IDs. This is meant to stop exactly this problem by ensuring every single device has a unique ID across all controllers, but if the update process interrupted the proper registration, the system might be confused and trying to manage two virtual lights for one physical bulb. Before you go through the massive headache of resetting every single bulb, check your router settings for any aggressive client isolation features. Some mesh systems or updated routers tighten up DHCP leases or MAC filtering in a way that disrupts the local network communication between the Thread border router and the individual bulbs. If you have a dedicated Thread border router or hub, make sure its subnet and DNS settings haven't been altered by the main router update.

Try removing the problematic group of lights and adding them back in small batches, maybe two or three at a time, instead of all at once. This forced re-pairing ussually clears the internal cache in the main controller and forces a clean registration using the newer standards. If the issue persists across the whole house, it might be worth checking if there are any micro-firmware updates available for your main hub that weren't pushed automatically by the router's changes.

The behavior you describe regarding dropping connections, even when the power indicator is on, sometimes points toward power instability rather than a pure network protocol failure. When a large quantity of low power devices are running, expecially during periods of high network activity, a slight voltage fluctuation or marginal power supply within a cluster can cause the radio module to drop out, even if the bulb itself appears illuminated. I have seen this happen with older dimmers that attempt to pull significant current at startup. It might be beneficial to verify that the devices are connected to robust surge protected outlets or consider a small power conditioning unit for the most problematic fixture groups. This is usually a matter of reliable current delivery rather than an addressing issue with the Thread border router.

I’m not convinced that this whole thing is just a bug in the Endpoint Unique IDs or the Matter registration. Pinning the blame entirely on the protocol transition seems like overcomplicating what could be a much simpler RF issue. All these modern smart home setups create a massive amount of digital clutter, and I’ve seen enough Thread networks that the sheer volume of wireless traffic is actually causing interference. It’s less likely that the system is confused about which bulb it is and more probable that the signals are being drowned out by background noise or competing radio traffic nearby.

I tried unplugging everything for a few minutes to see if that would reset the lights, but it didn't fix the ones way across the house. The idea of taking the whole cluster off and adding it back on sounds like a huge chore, though. I also saw a mention of changing the router settings. If I mess with the router's settings, am I going to break the internet completly or is it just for the smart gadgets?

Update: I read the comments about RF stability and tried tweaking the router settings today. Specifically, I changed the broadcast channel in the admin panel. Since the power resets did not fix the far-away lights, I am hoping that optimizing the wireless interference will get the devices to handshake properly again.