How to: Carry on a scientific collaboration

I'm writing this at the suggestion of a commenter on my previous how-to post, who was specifically interested in experiment/theory interactions.  Collaborations, as a fundamentally personal endeavor, are as varied as the people who collaborate.  Over the years I have collaborated with a number of theorist colleagues as well as fellow experimentalists, and generally it's been a very positive set of experiences, both scientifically as well as personally.  The main recommendations I can make about collaboration:

• Discuss and plan the ground rules at the beginning.  How is the collaboration going to work?  Is this the sort of collaboration that requires regular discussions and updates?  Are physical samples being sent by one party to another?  Which people are going to be responsible for what tasks?  What are peoples' expectations of authorship (recognizing that occasionally work may take an unanticipated turn, and someone's contribution may grow or shrink along the way)?  Are there restrictions about the samples or data?  (For example, a materials grower might collaborate with person A and person B on different projects; it could be very awkward if person A took samples and then on the side started working on the same project as person B!)
• Collaborate with people who have a similar approach to research projects as you, in terms of rigor, timeliness, and seriousness.  This is true whether those people are your own group, or outside collaborators. 
• Make sure to understand what your collaborators are actually doing.  Collaborations are a chance for you to learn something, since presumably you're working with these people because they bring something to a project that you can't do your self.  Sometimes asking what might seem at first glance a silly or naive question can lead to discussion that is informative for everyone.
• Have realistic expectations.  On the sociological level, realize that no one is going to retool their entire research enterprise or retask several people for your sake.  On the scientific side, know what can and can't be done by your collaborators and their techniques.
• Be communicative.  Keep your collaborators in the loop and up to date on what's going on.  If there is a big delay on your end for some reason, let them know.  You'd want them to do the same.  If you have decided that you don't think the project is going to work, or it's not working as anticipated, bring this up and don't let it sit.
• Be a finisher.  The most successful grad students are the ones who actually finish tasks and projects.  In the same way, don't let things slide.  If your collaborator wants you to read through a draft, or you promised to get some data to them in time for some deadline, follow through.