I forgot to update this thread. I'm unable to replicate this MTU problem in a more controlled environment. My next step is to set up a device at my house more similarly to how my VPS works namely by installing QEMU with the KVM module enabled and installing OpenBSD on it. Then using vio(4) as the physical connection wg(4) uses.
For now ignore what I've said; or more accurately, test what MTU you should be using instead of using the default MTU. This is probably "obvious" advice no matter what you're doing, but I rarely do this. All I know thus far is my VPS which uses QEMU + KVM with OpenBSD 7.9-stable installed and vio(4) as the network connection causes premature fragmentation on the wg(4) packets when an IPv6 endpoint is used even though premature fragmentation doesn't occur when communicating directly with vio(4) (i.e., 1500 MTU is correct). I may temporarily install Linux over my OpenBSD installation on my VPS to see if this fragmentation occurs; if so, this may suggest the issue is with QEMU/KVM.
