Losing data is one of the most serious bugs. When I am using a DBMS in production, I have to rely on it 100%.
I believe the complains made could be real because MongoDB is highly optimized for speed. But, as long as there is no documented and maybe reproducible case, this post can't be taken for real.
I'm very skeptical of the lost data claims. People using MongoDB are writing new code. New code has bugs. When data is lost, it's certainly more convenient to claim 'the datastore ate it' than to admit you have a critical bug in your own code.
I agree. And this is why I like CouchDB's versioning. In similar cases we could track down unwanted deletes using previous versions of the document in question. Without those, it could easily be interpreted as "data loss".
