Global climate model (GCM) projections remain hindered by a persistent spread in cloud feedbacks. This is despite efforts to reduce this model uncertainty through a variety of emergent constraints (ECs) dating back to CMIP3. Much of the early work in this area focused on constraining global cloud feedback, and in turn climate sensitivity, with a number of these studies suggesting an important role for present-day biases in clouds or precipitation. However, many proposed ECs have been shown to struggle with out-of-sample testing (verification of a meaningful relationship using a different ensemble). Given these difficulties, there has been a recent shift to targeting specific regions or cloud regimes. Following this approach, we use three generations of GCMs to assess the potential value of climatological cloud and precipitation metrics for constraining regional cloud feedback. We find that precipitation-based metrics are largely inconsistent in their relevance to regional cloud feedback patterns across the different ensembles. Alternatively, there is greater consistency for certain cloud-related biases. We find the greatest robustness for Southern Hemisphere surface shortwave cloud radiative effect (SWCRE) across the Southern Ocean, which is strongly tied to future cloud feedback across the tropics (30°S-30°N) in all generations. Using this relationship in conjunction with CERES data, we produce an EC on tropical cloud feedback. This EC suggests a tropical mean cloud feedback of 0.28 ± 0.35 W/m2/K, weaker than the unconstrained central estimate of 0.54 ± 0.5 W/m2/K. Lastly, we show the key role that low clouds play in driving this emergent relationship and illustrate why several previously proposed ECs struggle with out-of-sample testing.