What a routing protocol buys

A routing protocol replaces "routes typed in by a person" with "routes worked out by the routers, continuously, by talking to each other". That conversation is the whole point, and it pays off in three ways.

1. Failure detection, at both ends

Routers running a protocol send each other messages all the time: hellos in OSPF, keepalives in BGP, and BFD packets when you want speed. When they stop arriving, the routers on both ends know the path is gone, whatever the interface lights say, and recompute. That's exactly the gap the experiment exposed: static routes only know about their own interface.

2. Scale: the work grows with the network, not faster

With static routing, every router needs a route to every network it doesn't directly connect to. Ten sites with five subnets each means each router needs 45 static routes, 450 across the network, and adding one subnet means touching nine routers. Add a backup path and you double it. Every one of those lines is a chance for a typo that only shows up when something fails.

With a routing protocol, each router announces its own networks once, and the protocol tells everyone else. Adding a subnet is one change, on one router.

3. Knowledge you don't have

Inside your own network, you could in principle type every route. On the internet you can't: there are around a million IPv4 prefixes, held by tens of thousands of networks, changing every minute. The only way to know how to reach them is for those networks to tell you, and for you to tell them how to reach you. That conversation is BGP.

What it costs

None of this is free. A routing protocol is another moving part: timers, neighbours, policy, and new ways to fail, such as accepting a route you shouldn't or announcing one you mustn't. The rest of these lessons are largely about controlling those costs: filtering what you accept and send, protecting sessions, and monitoring them. The trade is worth it once any of the three forces applies; before that, it isn't.