[Spivak]

Not really, you're just trading one kind of downtime for another. You're not gonna get much meaningful improvement unless you implement actual redundancy. Docker makes this a giant PITA because they want you to use docker.io. It should be as easy as adding a second hosted registry or an additional private registry to the Docker search path (and it is that easy on Red Hat builds of Docker) but the official builds don't support it.

[nsonha]

I trade multiple points of failure to a single one. If you have redundancy for docker but your CI or cloud provider is down it would be useless too.