My first thought was: how come the frame isn't just dropped due to a bad checksum? Then I remembered a piece of Ethernet test hardware I worked with. It had a

My first thought was: how come the frame isn't just dropped due to a bad checksum? Then I remembered a piece of Ethernet test hardware I worked with. It had a bad SRAM chip or something which would result in all of the payloads having a stuck bit every 8 bytes. However, the FCS was calculated _after_ that, so it was still a valid frame. My guess is AT&T has some hardware that's recalculating the checksums after the frame has already been corrupted.