Why do my logs show "blocked" for traffic from a legitimate connection
Sometimes log entries will be present that, while labeled with the "Default deny" rule, look like they belong to legitimate traffic. The most common example is seeing a connection blocked involving a web server.
As also stated in this m0n0wall FAQ, this is likely due to a TCP FIN packet arriving after the connection's state has been removed. This happens because on occasion a packet will be lost, and the retransmits will be blocked because the firewall has already closed the connection.
It is harmless, and does not indicate an actual blocked connection. All stateful firewalls do this, though some don't generate log messages for this blocked traffic even if all blocked traffic is logged.
These blocked packets will occur even if rules exist which look as though they should match the traffic, such as an "Allow All" rule, as pass rules for TCP only allow TCP SYN packets to create a state. These rules assume TCP traffic with other flags will either be part of an existing state in the state table, or packets with spoofed TCP flags.
Blocked packets are also common for legitimate-looking traffic where routed networks and/or Multi-WAN are involved when Asymmetric Routing or other related causes are present in the network.
Clustering and Load Balancing
In a clustered environment, traffic arriving via the primary and leaving an internal interface can appear to be blocked on the secondary if the destination is a broadcast or multicast address like those used for Microsoft Network Load Balancing. The traffic appears to be blocked on the internal interface of the secondary from a public IP address source. Capturing the traffic on the secondary and inspecting the destination address in Wireshark will reveal the nature of the destination MAC address.
See Also: What are TCP Flags