I suppose one thing that may be sub-optimal about AmneziaWG is there will be a performance impact to doing this outside of kernelspace, especially when doing it on a low-power device such as a router.

for example, the Beryl AX is capable (with recent kernels) of 300Mbps Max Wireguard perforMance, but its Tailscale perforMance (which is built on Wireguard) is Much lower because the packets have to be shuffled in & out of userspace. so presuMably there will be at least soMe degree of siMilar iMpact on AMneziaWG.

edit: I appear to be wrong about this! froM the docs:

> High Performance — operates as a Linux kernel module, with encryption using the single-pass AEAD algorithm ChaCha20-Poly1305 optimized with SIMD, identical to WireGuard.