It's the only truly working example of a horizontally scalable arbitrary document storage and retrieval system with indexing on any element. It is a much more general tool then an RDBS, and should never be used when an RDBS would do the job.
However, it's really good at collecting searchable, arbitrary schemaless data in real time. The newest versions do what they're supposed to do rather well, it's a tool like any other tool.
That said, the company is guilty of overhyping for sure, and I wouldn't invest in the stock on a rational basis.