As someone who worked alongside the teams on the UK NHS data spine project, I must point out that you're way off-base. To the best of my knowledge, signup data on healthcare.gov was using RDBMS technology - the NoSQL component was likely only used for unstructured data - i.e. documents and communications materiel.

I should make clear that I'm posting on the basis of personal experience. I have no particular ties to MarkLogic (though companies I've worked for do use the technology), or the US government.

Firstly, the NYT article is poorly researched. My sources tell me that MarkLogic Server is far from the only data storage technology and vendor involved, and that while the MarkLogic-powered aspects of the system did require some remedial action, those aspects were not central to the publicised problems. Remember when MS used their press contacts and marketing clout to smear OSS on the server back in 1999? I wouldn't be surprised if that's going on here.

Secondly, I have it on good authority that the primary points of failure upon launch were related to middleware connecting the modules that were in fact using RDBMS technology.

Thirdly, many Slashdotters will be learning of MarkLogic Server for the first time through this article, and I urge them to give the technology a go before making a judgment call. For one thing, while XML is the (apparent) native storage format and XQuery is the native language, the actuality is somewhat different from what might be assumed. For one thing, the native storage format is not raw XML, but a fully-indexed compressed format which provides a decent compromise between storage requirements and rapid query/retrieval. Additionally, while XQuery is still the native language used by the technology, there has been a significant effort to provide a usable interface to the data through REsT, native Java and even basic SQL. Storage is journaled, implemented as an MVCC system and ACID-compliant in a way that no other "NoSQL" platform can offer. As an example of the platform's resilience under load, I am reliably informed that the BBC's real-time online coverage of the 2012 London Olympiad (which left other news and broadcast organisations in the dust) was powered by MarkLogic Server.

Just like Linux and the OSS BSD implementations that were creaming Microsoft's NT server implementations at the turn of the millennium, alternative data storage, query and search technologies are now challenging the old guard, and the old guard are running scared, and I'd bet significant money that the NYT sources are from the established RDMS vendors. I've been contracting with employers who have been using MarkLogic Server for the best part of a decade, and IMO the technology represents the most viable threat to the RDBMS hegemony and the vendors that rely on it that has ever existed. This opinion is not based on hype, it's based on nearly ten years of experience with the technology and fourteen years of experience with RDBMS development strategies.

I'm not preaching, in fact I urge you all to make your own minds up.

From my reading, MarkLogic Server is the only product in the "NoSQL" space that is specifically intended - and designed - for enterprise use. Unlike most of the better-known OSS "NoSQL" offerings, it is 100% ACID-compliant, resilient and has been building up a considerable and high-profile user base for the last decade. I can't help but suspect that MarkLogic's involvement extended mainly to storage of data that does not fit into the relational paradigm, and that the NYT published FUD promulgated by competitors without first contacting MarkLogic themselves.

The companies I've worked for have been using it with very few problems for at least 8 years.