I think it was late 2012 or early 2013 when I was in India (Hyderabad) at a time when an earthquake in the Mediterranean Sea cut I think it was SEA-WE-ME 3. The internet was terrible for the next two weeks, until the cable was repaired: almost half the internet flat-out didn’t work, including most USA hosts. I have no idea why communication between India and the USA, which should head east, was affected by a break to the west in the Mediterranean. I do know that local ISPs often have fairly dodgy peering arrangements.
My workaround was to tunnel via my own VPS in Singapore, as I could connect to it and but I was using OpenVPN back then and performance was pretty terrible. (Now if I want such tunnelling I use WireGuard, and it’s much better.)
We'd really need traceroutes (ideally bidirectional) from before, during, and after the break to diagnose your issues :P
But, even if your traffic was going east, with the broken cable to the east, there might be a lot more traffic going east (or coming from the east), and that could cause a lot of breakage.
For better or worse (mostly for worse), BGP doesn't propagate capacity of links, so it doesn't matter if there are alternate routes, if the overloaded route has the most desirable advertisement, it gets the traffic even if most of the packets are dropped into the sea.
My workaround was to tunnel via my own VPS in Singapore, as I could connect to it and but I was using OpenVPN back then and performance was pretty terrible. (Now if I want such tunnelling I use WireGuard, and it’s much better.)