With stuff like that you're still faced with a problem. What if a
customer has been directed to a server that is working fine to begin with,
and then dies. The 3DNS stuff will make the changes to avoid the server,
but the customer's browser won't pick up those DNS changes and they'll be
left hanging. Also, you'd have to keep the DNS TTL low or else you have
to deal with the DNS info being caching on ISP server. To further
complicate things the TTL is only the minimum required time to live. I've
discovered instances where cached DNS information will be stored for
weeks, even though the authoritative DNS TTL was originally set to a day.
Our only alternative is to eliminate every single-point failure with stuff
like high availability clustering, redundant feeds, battery backups,
nuclear reactors, physical separate sites on different planets, etc. 
(Pardon me, it's 2:00am and I'm getting punching)