Restarting juju machine unit process causes fan-network lxcs to lose IPs

Hi,

Using Juju 2.8.5. After upgrading to 2.8.6 (both controller and models), I've realized all my LXCs lost their IPs. Those are fan-networking LXCs.

Looking into this issue, I've realized that fan-networking addresses are set using a dhcp server.
I've realized that, after upgrading; that dhcp server will disappear and containers will fail with lost leases:

Nov 6 15:43:45 juju-cf278f-0-lxd-2 systemd-networkd[190]: eth0: DHCP lease lost

I've noticed that also happens after restarting machine's service.
The reason is that, although dnsmasq-dhcp process PPID is 1, on systemd, it shows as subordinate to jujud-machine:

             ├─jujud-machine-0.service
             │ ├─10808 bash /etc/systemd/system/jujud-machine-0-exec-start.sh
             │ ├─10819 /var/lib/juju/tools/machine-0/jujud machine --data-dir /var/lib/juju --machine-id 0 --debug
             │ └─11573 dnsmasq -u dnsmasq --strict-order --bind-interfaces --pid-file=/run/ubuntu-fan/dnsmasq-fan-252.pid --conf-file= --listen-address 252.0.53.65 --dhcp-range 252.0.53.66,252.0.53.78 -- ......

I can see on juju logs that it does run "fanctl up -a" after service restart:
/var/log/juju/machine-0.log:2020-11-09 10:05:42 DEBUG juju.worker.fanconfigurer fanconfigurer.go:69 Launched fanctl up -a - result /usr/sbin/fanctl: /etc/network/fan:40 10.192.0.0/12 252.0.0.0/8 -- failed err=1

The only way I can see now to bring the dhcp up is to manually run the same dhcp command.