Newton's Enigma?

Really, none of the above. It may increase or remain unchanged. I'm not sure if it can decrease. On average, and at sufficient distance, it's ogous to approaching a solid, spherical body of the same radius and average density; i.e., unchanged. But with a cluster you can be very close to the CG of one of its component bodies while remaining at a 1-radius distance from the composite CG. The inverse square law can have a large effect here, producing more g than you would experience on the surface of the equivalent solid body if the ratio of material density to average density is large enough. Basically the 2-body approximation fails and it's a multibody situation. Whether the result is an increase or a decrease in g depends on particulars of the conditions.