Over the past year we have from scratch built a significant web-application with MongoDB. We also have a social graph. We build infinite-scroll activity feeds on the fly. We handle multiple writes per second and 10s of reads per second. We have hundreds of thousands of users. We have 100 million documents, growing nicely. We've forced square pegs into round holes on occasion, but nothing we were too surprised about.

It seems like you walked into to your database technology choice with your eyes fully shut. Given even a modicum of of preparation - e.g. reading the MongoDB documentation - you would have asserted the social graph use case to be a challenging one with a data-store in which relations are unnatural.

Then - because you realized you may have a sub-optimal solution, you optimize it by changing technology. And then decide to join this ridiculous anti-MongoDB internet bandwagon.

For somebody who builds "4-6 web-applications per year" and has deployed "most of the data-stores you've heard about - and some you haven't" this seems surprising.

Or perhaps not, actually.