From what I understand, this means that everything in 192.168.2.0/24 will still be routed through the bridge (because /24 is smaller than /16), but everything else in 192.168.0.0/16 will be routed through the 212-Wireguard interface.
Is that correct?
The reason this came up is because everytime I enter a site I forget to add a local route to each router to the new site. And, I don’t know if this is accurate, but maybe fewer routers will improve the performance of the hEX?
Yes. It would work in exactly the same manner as versus 0.0.0.0/0 via your upstream ISP gateway … 192.168.2.0/24 has longer subnet mask and has thus priority over the 0.0.0.0/0 when deciding where to send packets for destination IP address e.g. 192.168.2.200.
The only problem with too short subnet mask is that some bogus traffic might get routed towards 212-wireguard interface and it has to dealt with there instead of hitting some black hole … and handling with “invalid” traffic inside wireguard process/driver is most probably pricier than dealing with it on pure routing layer.
With lots of sites you should consider using a protocol for dynamic routing, like OSPF. That way your routers will learn the routes automatically from each other, no more forgotten or borked configuration.
It depends somewhat where you are starting from. For a few routers in one area without route redistribution it is not too bad, but it assumes you understand connected routes, how static routing works, e.g. how routes are chosen, what netmasks are, etc. If you do decide to go this way, having something link GNS3 or EVE-NG to “simulate” multiple routers makes playing/learning a lot easier. There are some Mikro Tips by Druvis on setting up GNS3. The Network Berg also has videos about the advantages of having a virtualized lab.
See this post for some pointers to OSPF resources. I would start with the first two videos from The Network Trip’s OSPF playlist, then the one from The Network Berg.