It's a recently open sourced database by Yandex. It powers their web analytics backend. CloudFlare is already using it for their DNS analytics.
Their docs also talk about it a lot.
My approach breaks down if you are heavily updating your arrays though. On the mailing list Tom Lane suggested I store that kind of data outside the database. But of course all software development is an attempt to have your cake and eat it too, so I wrote an extension that is a bit like cstore_fdw, but maybe more extreme: you can store/load/append a float array as an ordinary file. It has similar operational disadvantages to cstore_fdw (no replication, no pg_dump, etc.) but performs faster. On the other hand it is definitely not battle-tested and still likely to change a little. (I need to improve the crash story and may add compression.) If you decide to use it, let me know!
Oh another limitation of keeping everything in an array (whether with floatfile or not) is that Postgres has a 1 GB limit on most RAM allocations. For example you could try running array_agg on a giant table and you might get a message about `Exception invalid memory alloc request size 1073741824`. I don't think cstore_fdw has that problem.
I also believe that the main use case is for the compression of data more than the performance benefits.