GRE Tunnel Issue

Has anyone run into an issue with GRE tunnels on a BIG-IP? I have a few setup running into a TGW in AWS and something seems to break them. Config change, Module change, ?? I haven't been able to pin down an exact trigger. Sometimes I could failover and have the tunnels on the other HA member work fine and failing back would results in tunnels going down again. (The tunnels are unique to each BIG-IP) They start responding with ICMP protocol 47 unavailable. Once this happens a reboot doesn't seem to fix it. If I tear down the BIG-IP and rebuild it, I can keep them working again for X amount of time before the cycle repeats. Self-IPs are open to the protocol, also tried allow all for a bit. No NATs involved with underlay IPs.