Why is XGMII little endian while the network byte order is big endian (RFC1700)?
Please realize that ethernet and IP were created by different people without knowledge of what the other was doing, and are today maintained by different organizations (ethernet=IEEE, IP=IETF). When Bob Metcalfe, et al. were creating ethernet, they were not working with Vint Cerf, et al. who were creating IP. These were two completely separate efforts. When ethernet was created, it was not at all clear that IP would become dominant, and ethernet was just one of several LAN technologies that was vying for attention. IP didn't get on ethernet for a relatively long period of time because IP was originally connecting universities across the country using WAN technologies.
Eventually, the market forces preferred the two technologies, and they ended up the dominant technologies in their respective niches, but, although they were both conceived (1970s) and born around the same time period (ethernet in 1980, and RFC 791 in 1981), they actually didn't have much to do with each other until the 1990s, and they are still maintained by completely separate standards organizations.
Ethernet has always sent the low-order bits of the high-order bytes first. That is simply the way it was designed to place bits on the wire. Other physical protocols can do it differently, and it has no effect on the upper-layer protocols any more than something like the symbol encoding used on the wire. The IETF settled on the network byte order to make sure that, regardless of the processor used in a host, the bytes on a network were consistently sent and received in the same order for use by the network protocols.