Last week while I was on vacation, my Microsoft colleague Steve Howard posted another detailed article on an aspect of SSIS that's frequently misunderstood - restarting SSIS packages after a failure. Steve is often asked to explain and demonstrate this feature during visits to customer sites, so he wrote it up on the TechNet Wiki in great detail with screenshots:
Automatic Restart of SSIS packages after Failover or Failure
He even has a suggestion for the tricky case where you don't want a Foreach Loop container to restart from the very beginning.
Other resources by Steve:
Steve works for Microsoft Services as a Premier Field Engineer for SQL Server in northern California.
Thanks, Steve!
After I posted this yesterday, MVP Jamie Thomson pointed out a concern about the reliability of checkpoints, especially where containers are involved:
http://social.msdn.microsoft.com/forums/en-us/sqlintegrationservices/thread/4968FB85-AC96-4CF2-B27E-4F23BAD2482B