In passing, Dare Obasanjo makes a very good point (talking about the inapplicability of RDBMS systems to solve very-large-scale requirements):
Luckily, there are only a handful of companies and Web services in the world that need to operate at that scale.
While I believe “a handful” is a little too restrictive, I’m pretty sure that for the vast majority of applications, an RDBMS is the right choice (if simple file-based store isn’t sufficient).
I think a number of these ideas also apply to maintainability, in particular the ability to evolve components more independently than tends to be the case in the IT data centers I am familiar with (and which seem to be the norm, from comparing notes).