Update: Oops, I meant TTL.. not ARP :D
The reason being is most users connect to Facebook or Instagram DAILY... which means their TTL cache would still be applied...
This happened either several days ago and users are just now starting to be affected as TTL cache expires... or something more deep. Actual routing problems.
This could be the start of the take down and the 10 days.. I hope all of us are prepared if so.
Update: DNS is not the cause, but FB's DNS server is a victim.
TL;DR, Their BGP AS "disappeared" from the Internet. This is extraordinarily rare, so rare that I've never seen it happen on a production network, unless it was a deliberate decision (like consolidating companies under a single AS).
For the normies, BGP routing is like how Verizon knows how to connect a call from your phone to your buddy's phone on AT&T's network. You know your buddy's phone number, but that call isn't going to get through. This is equivalent to Verizon not knowing AT&T even exists.
The BGP "issue" (I'm not saying failure because I don't think it failed) caused their DNS servers to become unreachable from the Internet, and since they run a very small DNS TTL, those IPv4 entries have aged out. What's interesting is that their IPv6 entries are still valid, but they're not native IPv6 end-to-end, so routing still breaks at IPv4-enabled segments.
There were no changes outside of the ordinary to their DNS in the last year. Also no routing changes, and I checked to see if there was a nexus between their networks and the DoD space that migrated out to Florida last year- there isn't. No planned outages scheduled, either.
While a route disappearing from BGP is common, an entire AS is not. BGP poisoning is a standard attack, but it doesn't make an AS "go away". This is more insidious.
My opinion, and not financial advice b/c I'm a retarded crayon eater, if this wasn't a US Government shutdown, it certainly is something in Russia's wheelhouse.. and Zuck pissed off Putin.