When the tunnel is up but traffic is dead
A WireGuard handshake proves that peers can exchange encrypted control traffic. It does not prove that application packets have a working path.
The handshake is only one layer
WireGuard can report a recent handshake while ordinary traffic still disappears. That is not a contradiction: the handshake confirms peer identity, key agreement, and UDP reachability between endpoints. It says nothing about whether an inner packet is routed, accepted by the host, forwarded, or allowed back.
That distinction changes the investigation. Regenerating keys is low-value once handshakes are current. The useful question becomes: where does the first inner packet stop?
Follow the packet in both directions
Start with routes and AllowedIPs on both peers. AllowedIPs acts as both a routing selector and an identity boundary, so an apparently reasonable route can still disagree with what the peer is permitted to send. Then capture on the physical and WireGuard interfaces to separate encapsulation problems from inner-path problems.
If the packet reaches the destination host, inspect local-input and forwarding policy, interface-specific nftables rules, and whether forwarding is enabled. For asymmetric paths, reverse-path filtering can reject valid-looking traffic before a firewall log ever explains it. The return route deserves the same attention as the forward route.
A repeatable debugging order
Verify endpoint reachability and the handshake, inspect routes and policy rules, capture the encrypted and decrypted packet, check host firewall counters, examine reverse-path filtering, and finally prove the reply path. This order keeps each test tied to a layer and avoids treating a healthy tunnel as proof of a healthy network.