I have two ER8411s in a LAN-LAN L2TP VPN configuration. The server has a public IP, the client is behind a NAT. The VPN is in a LAN-LAN configuration with route mode. The server router has a LAN IP of 172.18.0.1 and the client router has a LAN IP of 172.19.0.1.

From the client subnet, I can access anything on the server subnet as expected.

From the server subnet, I can intermittently access anything on the client subnet. Here's an example of a `ping` from a device on the server subnet to the client router:

```

$ ping 172.19.0.1

PING 172.19.0.1 (172.19.0.1) 56(84) bytes of data.
64 bytes from 172.19.0.1: icmp_seq=27 ttl=62 time=8.17 ms
64 bytes from 172.19.0.1: icmp_seq=28 ttl=62 time=8.92 ms
64 bytes from 172.19.0.1: icmp_seq=29 ttl=62 time=8.14 ms
64 bytes from 172.19.0.1: icmp_seq=30 ttl=62 time=7.61 ms
64 bytes from 172.19.0.1: icmp_seq=31 ttl=62 time=7.82 ms
64 bytes from 172.19.0.1: icmp_seq=32 ttl=62 time=8.39 ms
64 bytes from 172.19.0.1: icmp_seq=33 ttl=62 time=8.39 ms
64 bytes from 172.19.0.1: icmp_seq=34 ttl=62 time=8.94 ms
64 bytes from 172.19.0.1: icmp_seq=35 ttl=62 time=8.46 ms
64 bytes from 172.19.0.1: icmp_seq=60 ttl=62 time=192 ms
64 bytes from 172.19.0.1: icmp_seq=61 ttl=62 time=7.54 ms
64 bytes from 172.19.0.1: icmp_seq=62 ttl=62 time=8.26 ms
64 bytes from 172.19.0.1: icmp_seq=63 ttl=62 time=8.40 ms
64 bytes from 172.19.0.1: icmp_seq=64 ttl=62 time=8.54 ms
64 bytes from 172.19.0.1: icmp_seq=65 ttl=62 time=10.4 ms
64 bytes from 172.19.0.1: icmp_seq=66 ttl=62 time=8.57 ms
64 bytes from 172.19.0.1: icmp_seq=67 ttl=62 time=8.05 ms
64 bytes from 172.19.0.1: icmp_seq=68 ttl=62 time=7.47 ms
64 bytes from 172.19.0.1: icmp_seq=93 ttl=62 time=384 ms
64 bytes from 172.19.0.1: icmp_seq=94 ttl=62 time=8.39 ms
64 bytes from 172.19.0.1: icmp_seq=95 ttl=62 time=8.35 ms
64 bytes from 172.19.0.1: icmp_seq=96 ttl=62 time=9.32 ms
64 bytes from 172.19.0.1: icmp_seq=97 ttl=62 time=8.03 ms
64 bytes from 172.19.0.1: icmp_seq=98 ttl=62 time=8.22 ms
64 bytes from 172.19.0.1: icmp_seq=99 ttl=62 time=8.34 ms
64 bytes from 172.19.0.1: icmp_seq=100 ttl=62 time=7.72 ms
64 bytes from 172.19.0.1: icmp_seq=101 ttl=62 time=8.30 ms
64 bytes from 172.19.0.1: icmp_seq=138 ttl=62 time=2112 ms
64 bytes from 172.19.0.1: icmp_seq=139 ttl=62 time=1088 ms
64 bytes from 172.19.0.1: icmp_seq=140 ttl=62 time=64.3 ms
64 bytes from 172.19.0.1: icmp_seq=141 ttl=62 time=7.76 ms
64 bytes from 172.19.0.1: icmp_seq=142 ttl=62 time=8.53 ms
64 bytes from 172.19.0.1: icmp_seq=143 ttl=62 time=8.54 ms
64 bytes from 172.19.0.1: icmp_seq=144 ttl=62 time=7.71 ms
64 bytes from 172.19.0.1: icmp_seq=145 ttl=62 time=7.66 ms
64 bytes from 172.19.0.1: icmp_seq=146 ttl=62 time=8.36 ms
64 bytes from 172.19.0.1: icmp_seq=147 ttl=62 time=8.98 ms
64 bytes from 172.19.0.1: icmp_seq=148 ttl=62 time=8.52 ms
64 bytes from 172.19.0.1: icmp_seq=173 ttl=62 time=256 ms
64 bytes from 172.19.0.1: icmp_seq=174 ttl=62 time=8.53 ms
64 bytes from 172.19.0.1: icmp_seq=175 ttl=62 time=7.71 ms
64 bytes from 172.19.0.1: icmp_seq=176 ttl=62 time=8.33 ms
64 bytes from 172.19.0.1: icmp_seq=177 ttl=62 time=8.66 ms
64 bytes from 172.19.0.1: icmp_seq=178 ttl=62 time=9.10 ms
64 bytes from 172.19.0.1: icmp_seq=179 ttl=62 time=7.97 ms
64 bytes from 172.19.0.1: icmp_seq=180 ttl=62 time=7.73 ms
64 bytes from 172.19.0.1: icmp_seq=181 ttl=62 time=8.40 ms
64 bytes from 172.19.0.1: icmp_seq=203 ttl=62 time=3009 ms
64 bytes from 172.19.0.1: icmp_seq=204 ttl=62 time=1985 ms
64 bytes from 172.19.0.1: icmp_seq=205 ttl=62 time=961 ms
64 bytes from 172.19.0.1: icmp_seq=206 ttl=62 time=7.75 ms
64 bytes from 172.19.0.1: icmp_seq=207 ttl=62 time=7.62 ms
64 bytes from 172.19.0.1: icmp_seq=208 ttl=62 time=9.13 ms
64 bytes from 172.19.0.1: icmp_seq=209 ttl=62 time=8.60 ms
64 bytes from 172.19.0.1: icmp_seq=210 ttl=62 time=8.19 ms
64 bytes from 172.19.0.1: icmp_seq=211 ttl=62 time=7.82 ms

```

You can see that sequences 68-92 failed, 102-137 failed, 182-202 failed, etc.

It seems that server->client pings succeed when I am accessing resources on the server subnet from the client subnet, and the pings fail when there isn't any client->server traffic. Almost like the client->server traffic creates a temporary routing rule on the server side, that temporarily allows the server->client traffic?

It's very odd behaviour, but I need to fix it so that devices on the server subnet can consistently access devices on the client subnet.