But, what I haven't found is anyone talking about using the ndbclient API in large scary production environments. Is there anyone out there really hammering it?
Infact, a Google Search for "ndbclient" only returns 111 results right now, none of which look too great.
It seems like it could be a great solution instead of using application level partitioning and replicating Berkeley DB in some use cases. I don't mind being given a low level API and having to build my application level logic on top... It's how Sleepycat has worked for 20 years.