Seems GitHub moves to Route53 for DNS as problems appear to be on/off most of the day.
GitHub System Status
I didn't even consider some zones have pretty high NS TTL so it can take some time to even attempt to migrate issues like this.
Long TTLs will make any attempt to migrate away from the issue difficult, but at the same time they should also allow all of the other DNS servers to cache the results - if my ISP has cached the lookup for whatever name it shouldn't matter that Dyn is getting attacked. Considering the very distributed cache-friendly* nature of DNS I'm quite surprised by the impact this had.
* When I say 'cache-friendly', I mean more that is has extensive caching built right into the protocol, to the point that the caching almost causes more problems than it solves, so people turn down TTL's to work around the cache issues, which then makes them vulnerable to an attack like this.