Or it's the opposite, where the slowest possible retrieval time is the intended effect, as is the basis of many cryptographic algorithms.

Or it's neither, and the intended effect is zero variation in the retrieval time, as when trying to avoid leaking secrets via timing attacks.

(Or I guess, more generally, the intended effect is zero correlation between the information and the time it takes to retrieve it. If retrieval time were completely random, it would achieve the goal, but it wouldn't have zero variation.)