Checkpoint files are very useful mechanisms for recovering failed packages and I have talked about them quite a bit before, however, they do have one or two limitations.
One thing that people typically want to do is always execute a particular task regardless of whether a checkpoint file exists or not. There may be a number of reasons for this - the most typical scenario is thus:
Data is staged from a source system and downstream some errors in that data cause another task to fail. Later when you fix the bad data in the source system and re-execute the package you don't want only the failed task to execute first - you want the task that fetches the data from source to execute again as well.
Up until now I didn't think that was possible but recently David Noor from the SSIS product team explained how to do it in this thread. Rather than repeat verbatim what David said I have instead recorded a video that demos exactly how you can configure this. The video is below but if for whatever reason you can't see it, go here to see it on Youtube instead.
I have attached the package that I built in the video.
Comments are welcomed.