Tailscale node advertising routes goes down, whole local network disappears for another node on same network

Tailscale version: 1.30.2 & 1.22.2
Your operating system & version: Windows & Synology

I run Tailscale on a couple of windows machines, a Raspberry Pi, a Synology NAS and a Linux VPS.

Initially I did it as I got cable internet, but had to have CGNAT. I pay $1/month for a hosted VPS on which I run NGINX and reverse proxy through Tailscale to my Raspberry Pi, where I host some websites.

I also installed Tailscale on my desktop, my laptop, my iPhone and my Synology NAS. On the NAS I allow routing to the entire 192.168.x.x network and on the Pi I allow use as an exit node.

I recently got a static IP so the setup was no longer needed. I also got a new router. It all worked fine but 24 hours later my laptop stopped talking to my router. And my Pi. My desktop didn’t work either. But they could talk to the internet (and to the router, by going out to the internet first with hairpin NAT). Initially I blamed the new router, reset it, changed the config, but noticed my phone (on which Tailscale was turned off) did connect to my router and my Pi. Weird. I then found that turning off Tailscale on the laptop sorted things out. And that Tailscale on the Synology NAS (advertising the subnet routing) had crashed.

Is it normal for the home network to become unreachable if the Tailscale node advertising routes is on the same home network and it goes down (but the home network stays up)? I have solved it by getting a static IP rather than CGNAT now, but is there a way to combat this problem when the node goes down? This is the second time it has crashed on my Synology in the last month (I didn’t notice it last time)- I would have moved the route to my Pi which has been more stable.

There’s a longstanding GitHub issue that is somewhat related. If Tailscale on the Synology was advertising a route to the home network, and Tailscale on the laptop was accepting the route (even though it has a direct connection to the same network), it is very possible for the laptop to try to route ‘home’ traffic through the VPN.

Currently the best solution is to just stop accepting routes on the laptop while it is connected to the home network, and start accepting routes again when you take it to a different location.

It may have been doing that all along, I don’t know. But when the Synology crashed it was still trying to route home traffic through the VPN and failing. I’m not sure why it didn’t notice that the NAS had failed. The shortcut icon had even disappeared from the NAS desktop (as it did previously).

I spent quite a while fiddling with the router and I could easily have been fooled into spending even more - I’m not sure now what caused me to turn off Tailscale and immediatrly solve the issue. Possibly because it had crashed once before not long ago and I had thought of moving the route to the Pi at the time (I had read reports of crashes on Synology, but not of my issue) and because fortunately it worked on my phone - I use several VPNs on the phone so don’t leave Tailscale on permanently like I do on my laptop/desktop. But it was 2:30am and I could have easily just plodded on going around in circles.

I have now removed the routes. Didn’t really need them anyway as I have other conventional VPNs with my static IP.