Better to ask about their minimum availability guarantee. Knowing that the servers can handle a DDOS attack is of little value if the service goes down for maintenance every day (just to give an example).
GitLab.com runs the same GitLab as our users and some upgrades require downtime. Other operations should not require downtime, we have multiple application servers and load balancers. The backend services are HA except the file storage until we complete the move to CephFS.
I'm not sure whether we've seriously considered Swift, but high bandwidth is important to improving our performance, one area in which Ceph seems to be the better choice [0].
