> Instead had the approach been ipv6 only from the start, with no dual-stack, having the OS transparently deal with sockets to ipv4 devices by converting to the ipv6 mapped address (:ffff:xxxxxxx), and thus eliminating the need for dual stack from the start, things would have moved far far faster.

This is, indeed, how dual-stack works; you open a PF_INET6 socket and use sockaddr_in6 addresses for everything, including IPv4 (which get mapped to ::ffff:/96 addresses). Been like that essentially forever. The “dual” in dual stack refers to the OS' stacks, not userspace.