We show that delaying fully-expansive proof reconstruction for non-interactive decision procedures can result in a more efficient workflow. In contrast with earlier work, our approach to postponed proof does not require making deep changes to the theorem prover.