The problem is that, devoid of user interaction, the wetness value will increase along the time axis. Some burgers are quite dry and will increase only moderately. Others, however, will leak out the sides, leaving the bottom bun soaking, and actively discarding liquified filling at worst.
The user interacts with the burger in such a manner as to reduce its wetness rating. This often involves at least rotation of the burger on one axis, and often is elevated to non-symmetric consumption and rotation about multiple axes. Eating time (volume minimisation efforts) may be traded off against the benefits of filling-redistribution for extreme cases.
The exercise essentially boils down to finding a continuous minimum along the time axis, for a particular burger, whilst manipulating the user interaction variable. Note that while the limited ability to look ahead means that what produces a minimum wetness in the short term is not guaranteed to enable minimum wetness in the moderate or long term, gains made in the short term often convey significant advantages later on.
One thing to keep in mind is that all burgers are different, and excellent strategies for one burger are not necessarily excellent strategies for the next - or even good strategies at all. This was summed up by the No Free Lunch theorem of Wolpert and Macready in search and optimisation. Indeed, the theorem states that the strategy you employ that actually minimizes the wetness of one burger will in fact be the very strategy that maximizes the wetness for some other burger.
Think about that the next time your absentmindedness leaves you with a burger that's composed of six breadcrumbs floating on a sea of mayonnaise that drowns two lettuce leaves. There's no free lunch when dry-cleaning is involved.
