Hey Pedro, We recently made some updates to our Lazarus service on the Backend that should address Prefect Tasks getting stuck in a running state, this may help address what you're seeing on the ECS side but definitely let us know if you continue to see strange behavior. Otherwise you might check out this document from AWS, for image clean up suggestions for the ECS tasks
https://docs.aws.amazon.com/AmazonECS/latest/developerguide/automated_image_cleanup.html