Wouldn’t you build in a failsafe that bypasses Fastly and sends traffic to your own servers in the case of this kind of outage? Or outages are so rare that it’s not worth the trouble?
The number of serious CDN outages in the world are incredibly rare.
In fact, you can probably remember most of them if you were given dates.
Plus, going around the CDN can be very complex (depending on the type of content), very expensive (all of a sudden you have a massive data out network traffic that didn't exist previously), and not guaranteed to work (DNS updates can take longer to get to everyone than the actual CDN outage lasts).
There are places where it is worth it and useful, but for a lot of the sites listed it's not useful.
That's the fallback, but the original stack is not designed with the volume of traffic in mind. So it gets overwhelmed very quickly and makes the website practically unavailable.
> Or outages are so rare that it’s not worth the trouble?
This, I can't remember the last Fastly outage in this dimension, so the time spent on setting up a secondary server serving your assets is probably not really worth it for small-medium companies. Although i'd think otherwise for a company like Shopify.
Many sites do this; Amazon's failed over to their own servers for images for me, it appears. It typically just takes some human intervention, I suspect.
Wouldn’t you build in a failsafe that bypasses Fastly and sends traffic to your own servers in the case of this kind of outage? Or outages are so rare that it’s not worth the trouble?