However, you both seem to be missing the point I'm making concerning the empirical evidence this incident highlights. I see no point in repeating myself or the raw facts I've presented, but aside from pc's speculation that there were multiple problems coincidentally involved, neither of you have explained why huge swaths of the Internet could not reach my server's fixed IP address while others could. You seem to be saying that DNS was not the issue, and if so you haven't explained what took place. It seems clear that DNS was the entire reason for the incorrect routing despite IP addressing being correct on source and destination.
That happened to me aswell recently. The cause of this was probably the
"peering" of your datacenter and a technical issue there.
Technically, your datacenter has several "partner" internet providers,
such as Telekom, AT&T and what not. All of them have their own "peer" to
the datacenter and to that peer corresponds a set of "routes" that this
peer handles. For instance, someone from telekom would reach your server
from the telekom peering while vodafone customers would reach it via the
vodafon peering. If one of them fails (due to hardware or configuration
errors, or maybe even a malicious BGP (border gateway protocol)
package), those customers from that peering couldn't reach your server.
They also can't really find a route to your machine via another peering
(e.g. telekom, vodafon, what not) because either there are not other
peerings for the destination network, or the "costs" for that route are
too high (in network design, you want to prevent packages from being
routed in a circle!).
Stuff like that is a little similar to so called "net-splits" as they
are known from IRC.