@pe1chl: It's generally true, but if this thing can be implemented as easily as authors claim:
WireGuard has been designed with ease-of-implementation and simplicity in mind. It is meant to be easily implemented in very few lines of code, and easily auditable for security vulnerabilities.
(even though "very few lines of code" sounds a little too optimistic), it might be worth to give it a higher priority. If implementing Wireguard would be easier than finishing OpenVPN implementation (I don't know, might be), I'd say to go for it. Not that it's a dream come true in complete package...
I have mixed feelings about roadwarrior use. It needs only single udp port (great) and even has some kind of roaming (I'm still not decided how much it helps). But inside config (addresses, routes) seems to be intentionally static-only. That's not great, because it means that it's not very usable when there's a lot of users and things can change. On the other hand, it's not much worse than what MikroTik's OpenVPN offers. For small SOHO use it could be good, as it seems to be otherwise quite easy to understand. Even working Windows client already exists.
For site to site, IPSec works great for me, but it's true that I do it mostly with static public addresses. When that's not available, Wireguard could work better. It should also have better performance on devices without HW acceleration. And it would provide interfaces for links, which would make it more clear for a lot of people than current tunnel-mode IPSec (I know about IPIP/GRE/EoIP inside IPSec, but it's extra step).