Debiasing Causal Inferences: Over and Beyond Suboptimal Sampling