The classic Veach Door scene was originally made to challenge the efficiency of all rendering algorithms, except for metropolis light transport. Only a small fraction of light emitted within a neighboring room passes through the opening in the slightly ajar door. At equal sample counts, our radiance-driven neural path guiding (NPG-Radiance) is better at finding paths through the door opening than the competing technique by Müller et al.  and is further outperformed - especially on the metal teapot - by incorporating the product (NPG-Product). Primary-sample-space path-sampling techniques offer little to no benefit over unidirectional path tracing with our NPS slightly outperforming PSSPS [Guo et al. 2018] with an equal number of samples.