One cannot serve two masters, but if the action space is convex, your masters report preferences honestly, and you take Pareto improvements, then you end up serving a linear combination of two masters which assigns nonzero weight to each. Instance of Complete Class Product™
0
11
38
0
1