Anyway, again, I agree with the underlying premise. It seems like overkill to need to have a fallback geocoding service, but I've use Geocod.io in the past and it seemed pretty cool. I have to say though, for a single application it seems like implementing 2 of every external service you use is a ridiculous thing to expect.
That's like saying deploying a service on more than one machine in different data centers in case one of them goes down is unrealistic. We've long past that and long recognized that it is not only realistic but necessary once you're on certain scale.
Are you sure you've read the article?
If you need a feature to work in the presence of down servers, you need to make your own fallback or perhaps cache some of the results locally.
I upvoted you because you make a good point about the fact that apps should handle network/service failures appropriately such that ideally it isn't a crisis but that doesn't make it an acceptable level of service.