I was very happy with Scalegrid hosted MongoDB for the 2 years a project ran on it. Prior to that, 1 year on a self-hosted cluster had a failed write due to a n

I was very happy with Scalegrid hosted MongoDB for the 2 years a project ran on it. Prior to that, 1 year on a self-hosted cluster had a failed write due to a network partition and I learned about single write masters and used CouchDB on my next "document-DB-appropriate" project. Atlas ended up being expensive and the service was a bit stand-offish, where Scalegrid's staff were super helpful etc... Their pricing was the primary factor. What do I define as "document-DB-appropriate"? Blobbed stuff where you want it all in the same document sized doses, and always the same dose, to the point where you feel ridiculous putting bits together in the same shape over and over and notice that you CRUD the same dose/shape all day... Objects... Don't get me wrong, there's times in computer data land when the first 128 bytes are the header and that the data is multiplexed in 32bit chunks with 24 bits zero-padded and this padding tells you "channel 2 is starting now buddy!" But SQL is certainly a viable option for many things and rather standard and known and supported and stuff...