Careful analyses are useful (and I've seen some well-reasoned debates in HN) but what strikes me is that clear-cut cases for using one tool vs another are rare, for web-dev back ends at least. Witness: the large-scale sites that use various imperfect db solutions. Yes, they sometimes involve some "hacks" to the canned db solutions, but who's found the perfect swiss-army-knife that needs no modification for any project.

What does strike me as significant is the mastery level of the db developer (whatever the db on which they have the greatest mastery/comfort) ...because projects sometimes "go south" or end up getting late-stage additions of unforeseen complexity.

Conclusion: I'm a db agnostic. Or maybe "poly-theist." There are several good competitors. None is perfect. Several are powerful, especially in the hands of an insightful and experienced db hacker. ( IMHO )