A team deploys a web application to an Amazon EC2 fleet with AWS CodeDeploy in-place deployments. During an incident, an engineer manually cleaned up disk space on a subset of the instances and deleted files under the CodeDeploy deployment root directory. Every deployment since then fails on exactly those instances during the ApplicationStop lifecycle event with "Script does not exist at specified location", even though the new revision bundle contains the script and the same revision deploys successfully to every other instance in the deployment group.
What explains the failure, and what is the most direct way to get deployments working again?