Monosynaptic tracing using rabies virus is an important technique in neuroscience, allowing brain-wide labeling of neurons directly presynaptic to a targeted neuronal population. A 2017 article reported the development of a noncytotoxic version-a major advance-based on attenuating the rabies virus by the addition of a destabilization domain to the C terminus of a viral protein. However, this modification did not appear to hinder the ability of the virus to spread between neurons. We analyzed two viruses provided by the authors and show here that both were mutants that had lost the intended modification, explaining the paper's paradoxical results. We then made a virus that actually did have the intended modification in at least the majority of virions and found that it did not spread efficiently under the conditions described in the original paper, namely, without an exogenous protease being expressed in order to remove the destabilization domain. We found that it did spread when the protease was supplied, although this also appeared to result in the deaths of most source cells by 3 wk postinjection. We conclude that the new approach is not robust but that it could become a viable technique given further optimization and validation.