This talk discusses the problem under what conditions two extensions of a Lorentzian manifold have to be the same at the boundary. After making precise the notion of two extensions agreeing at the boundary, we recall a classical example that shows that even under the assumption of analyticity of the extensions, uniqueness at the boundary is in general false — in stark contrast to the extension problem for functions on Euclidean space. We proceed by presenting a recent result that gives a necessary condition for two extensions with at least Lipschitz continuous metrics to agree at the boundary. Furthermore, we discuss the relation to a previous result by Chruściel and demonstrate a new non-uniqueness mechanism for extensions below Lipschitz regularity.