High Amount of Internal Traffic

I am running three virtualized x86 instances of Openwrt with the Zerotier package. I’m seeing some really weird behavior. I’m bridging my ZT interfaces to the internal br-lan bridge. When I run tcpdump in br-lan I get this:

13:43:02.395313 IP 192.168.20.1.29994 > 192.168.20.2.44862: UDP, length 332
13:43:02.395379 IP 192.168.20.1.29994 > 192.168.20.2.44862: UDP, length 1122
13:43:02.395425 IP 192.168.20.1.29994 > 192.168.20.2.44862: UDP, length 417
13:43:02.395447 IP 192.168.20.1.29994 > 192.168.20.2.44862: UDP, length 1432
13:43:02.395475 IP 192.168.20.1.29994 > 192.168.20.2.44862: UDP, length 95

Over and over. It literally creates some type of internal loop. It produces 150GB of traffic on the interface in literally minutes. None of the traffic is going out the WAN interface, it all stays internal. At the same time no traffic can travel across the vpn, but the peers seem to stay up. If I reboot all 3 nodes it seems to fix it for some ransom amount of time. When it is this weird condition, zerotier-cli shows the private addresses as peers and not the public addresses.

openwrt:~# zerotier-cli peers
200 peers

50e297b66c 1.8.4 LEAF 35 DIRECT 3 7340 192.168.20.2/44862
62f865ae71 - PLANET 231 DIRECT 7381 2200 50.7.252.138/9993
778cde7190 - PLANET 83 DIRECT 7381 2321 103.195.103.66/9993
b15644912e 1.8.9 LEAF 45 DIRECT 278 234 35.222.184.52/55005
cafe04eba9 - PLANET 247 DIRECT 7381 2226 84.17.53.155/9993
cafe9efeb9 - PLANET 8 DIRECT 278 270 104.194.8.134/9993
dbc48d2cdf 1.8.6 LEAF 16 DIRECT 3 16246 192.168.20.3/9993

Somehow it seems to have confused the internal and external addresses. It does not make sense that when I sniff on my lan bridge, that I am seeing the UDP VPN traffic. That should only appear on the outside wan interface. I’m at a loss. It seems like a loop, but I’m pretty sure that I’m not looped anywhere on my side, and I assume that zero-tier has internal loop prevention. Thanks for any help. Zerotier would be great if I can get this issue fixed!