wireguard seems to be crashing after adding a peer

[guest42406]

i was observing a weird behaviour today.
we are hosting an on-prem network with some vms running behind an opnsense.
on-prem wireguard is working flawless.
additionally we are running some services in a hetzner cloud environment.
like described here "https://community.hetzner.com/tutorials/vpc-with-wireguard-pfsense" i wanted to link on-prem and cloud, but instead of pfsense with an opnsense. i used the image provisioned from hetzner (23.7) and installed opnsense into an 3c/4gb ram machine.
after setting up i wanted to create the tunnel, but as soon as i add any peer to the config, wireguard seems to die. see the attached log file. somehow it cant configure "wg1", manual config via cli doesnt work either.
wg-instance is offline, no ip assignment.
additionally the server is very unresponsive, restart commands take some time and during boot "configuring wireguard" sometimes takes up to a minute.
the only way to get the wg-instance back up is to remove all the config, disable and reboot.
it always happen to die after i add a peer, with just the local instance set up, at least the instance is running and assigning an ip adress to the wg interface.
weird thing is, dmesg printed that "wg1" went down, ifconfig shows it as up.
hetzner has this sdn-thing going on to privately connect vservers, so i get a hop on the lan side assigned via dhcp. and i added an anti-lockout rule, as per default the lan route somehow takes precedence over the wan, even with "mark as upstream" set and higher priority.
i already tried to reinstall, change names, change ip-adresses and to add another wg interface.
also i tried the wireguard kmod and go implementation, same behaviour, different looks.
openvpn is not running, neither does ipsec.
current version running is 23.7.12
please let me know of anything additional needed.
help would be appreciated

[guest42406]

weird stuff is continuing today.
i managed to somehow get a second instance up, as i removed the first then, the second stopped working too and threw the same error messages as the first before.

also other stuff is going on, see the screenshots.
one instance is coming up with a random listening port, cli output cant assign it to a process.