These are basically flip sides of the same coin. Granting moderator for life allows independence from the community so that moderators can move in the direction that they feel is needed without having to worry about the immediately popularity of a decision, but it also removes a lot of control from the community itself.
On the other hand, confirmation votes ensure that the community maintains overall control, but makes it difficult or impossible to handle anything that is unpopular with the community (or perceived as possibly being that way), even if it really would be best for the community.
It also isn't a 100% either/or type situation. You can have hybrid situations where the community can override moderators on a case by case basis in certain cases (such as on StackExchange), cases where an independent group (possibly owners of the site) police moderators to make sure they aren't straying too far outside of community desires as a whole, or cases where moderators can be recalled, but require a larger amount of effort to remove than is required to grant it in the first place.
Each of these helps try to find a balance between the advantages and disadvantages of both extremes, thus tying the moderators to the general long-term will of the community, while making them relatively safe from short-term backlash for an unpopular, but necessary, decision.