There was a discrepancy between Bitcoin Core and btcd - that I noticed with bitcoinfuzz - about the rejection of IPv4-in-IPv6 when parsing an addrv2 message. btcd ended up implemented the same behavior while it raised a discussion about this not being mentioned in the BIP.
Rejecting this kind of addresses is important for canonical encoding. The whole point of this BIP is having exactly one representation for each network. Besided that, there are other reasons such as reachability accounting, defense against evasion/bypassing address‑type policy, etc.
That said, this PR changes BIP155 to document the behavior where IPv4 addresses embedded in IPv6 should be rejected.
