A valid IPv6 address can end with a IPv4 dot-decimal, such as: 2001:db8:4321::192.0.2.123 which effectively translates to 2001:db8:4321::c000:027b i.e. the 32 bits of the IPv4 address are simply mapped to the last two fields of the IPv6 address. When an SSH known_hosts file contains a line like: example.org,192.0.2.123,2001:db8:4321::192.0.2.123 ssh-ed25519 … then on a v6 connection, a new line is added to the known_hosts file, which is just the translated IPv6 address, e.g. 2001:db8:4321::c000:027b ssh-ed25519 … This creates redundant information, and it would be better if SSH recognised that the IPv6 address is already used and mapped to a host fingerprint.