The main things are removing the cellular connection and disabling the connection back to the Tesla services. Back in the old days you could pull the SIM card, root the center and driver’s displays, setup firewall rules to block traffic to/from the Tesla servers, and disable the VPN.
This is more difficult with the newer models. You can still pull the SIM, but would need to get creative for root access since it is a continuous game of whack-a-mole between the root methods and patches.
I set the VPN tunnel from the VPS to deny everything to the internal network by default, then put the services that need to be accessed on the allow list in the firewall. So the VPN endpoint from the VPS can only hit the very specific IPs/ports/protocols that were explicitly allowed. There is still the possibility of a compromise chain of VPS->service->container/VM->hypervisor->internal network access, but I feel comfortable with those layers.
You could also setup an IDS such as Snort to pick up on that exploit traffic between the services and internal VPN endpoint if extra security is necessary on top of fail2ban and log alerts on the VPS.
Depends what you want to do. They don’t require a network connection to operate as a vehicle. So if you don’t care about the remote app features (local ones such as lock/unlock still work over BLE), live traffic, streaming music or updates, then a network connection isn’t necessary.
If you do want any of those features, then you would need to either get root access to the gateway and infotainment systems to modify the endpoints or take over the C&C server (formerly named “mothership”) domains and certificates.