Within SSIS checkpoints, even by their naming, are confusing. Others have provided feedback about the properties it utilizes to decide if a task should be rerun on failure. But I think that it is even more of an issue when using containers. Take for instance a package with a few tasks, then a sequence container with few tasks within it, and then a few more tasks. Assume also that all items are connected sequentially with constraints, not to run in parallel. If a failure occurs on a task prior to or following the sequence container I could easily control if an item would be rerun. But if it failed within the container, then I would assume I could control what will be rerun by how I configure the FailPackageOnFailure and FailParentOnFailure of the container and the tasks. If I want to treat the tasks within the container as a set of tasks to always be rerun as a whole this is not possible. Based on having the FailParentOnFailure property it would make sense that setting this to true could be used to fail the container. And by not setting the FailPackageOnFailure for the tasks in the container it would not remember the completion of the tasks within. This is obviously not how it works. I know I can get this sort of functionality if I move the sequence container items into a subpackage with no checkpoint file.