I think the first means an intention to have operations continue relatively unimpeded, to the extent that the cost of the plan/fault-tolerance per period-of-time doesn't exceed (some percentage of) the likely loss over period-of-time without the plan/fault-tolerance. In effect, to prevent harm to the extent reasonable.

The second means an intention to reduce harm, but assumes prevention of harm is unreasonable from a cost perspective. It may be for events forseeable but so unlikely that the cost of testing is not reasonable. It also may therefore be the mobilization of very expensive resources to deal with incredibly expensive, but very rare, failure.

In that light, say, the New Orleans flood was a failure of the first type, followed by a failure of the second type.