We've been doing lots of work building out new cookbooks for our devops system, and although our production scripts are pinned to a known stable version, it seems like they ran a newer, unstable version instead which caused a key to be corrupted during a converge. We discovered why production scripts weren't pinned as expected, so we can avoid this in the future. Also if you're still reading this, maybe you want to join our team as a full-stack engineer ( jobs.lever.co/spark )