The reason I built this instead of just pointing at vendor docs is the case that actually breaks real deployments: pairing a policy-based peer with a route-based one. A route-based side (VTI/tunnel interface) defaults to a catch-all 0.0.0.0/0 ↔ 0.0.0.0/0 selector, since routing decides what enters the tunnel. A policy-based peer negotiates specific proxy-IDs per subnet pair, pulled straight from its ACL. Put those two together unmodified and phase 2 negotiation just fails — no useful error, just a tunnel that won’t come up.