To get our code to production what is left to do is to turn the “potentially” releasable product into a releasable product. To do this a number of activities are required: Which ones and how much effort they require varies greatly between types of organisation.
The fastest way to perform rollout activities is to do them within the development team and to automate as much of the process as possible. While I envy and respect teams that are mature in terms of test and deployment automation and operate in an organisational context that allows them to easily move code from environment to environment I find that this is very often not possible within larger organisations.
Often, among other activities, someone outside the team has to be instructed as to how to deploy code, other parts of the organisation have to be informed about the changes (eg. end users, operations and maintenance teams, etc), and manual regression testing has to be performed.
For the kind of large organisations Mike Cohn refers to in his description of a bank with COBOL code and no automated regression testing or large organisations new to Scrum and Agile I, too, find this is best done within a so-called release sprint.