I was hired to write the Etsy Middle Layer "Emid" server. They told me at one point that they were arguing about being able to find python developers and one of them went on Craigslist to settle it and mine was the first resume they saw. (That was personally kind of a let down. I just got lucky.) We had a phone interview. I mentioned that I used Dvorak and that clinched it. I worked from home for months before ever meeting the others.
The "emid" server was pointless: nearly three hundred api calls and fully 80% of them were bodyless wrappers. Just a decorator that did the real work, the function def, and a docstring to please the syntax (could have used a pass statement.) All the real "business logic" was in stored procedures in the DB. Yes, the DB was constantly on fire as we scaled up (our numbers were mind-blowing the whole time I was there. Up, up, up, and accelerating.) The idea was that emid would sit between the web servers and the db and provide some caching and other scaling support, but we never effectively used it. We actually did not turn on the caching until the investors forced an outside tech consultant on us to try to get some management of the horrible scaling fires that were constantly happening. So emid just sat there consuming resources. It had one good use though: when the system messed up (because the DB was almost-literally on fire) the ops/dba dude could complain to the CEO about "the emid problem" and pass the buck.
There's a lot of stories I could tell, but I don't want to "talk trash". Overall the users LOVED us and that's what really counts, eh?
Also, for what it's worth, the whole "Etsy was born in 2005 in a Brooklyn loft as a way for one of its co-founders to sell his handmade wooden goods." story, although it sounds great, is B.S. as far as I know. The original three were pretty close with the info (I do NOT know what "Etsy" stands for or means, don't ask) but they did let it drop that they had picked up the site from someone else who had originally made it but didn't want to do anything with it. Certainly the original three are responsible for it taking off like it did, them and dumb luck.
One highlight was that I got to work with Glyph (of Twisted networking engine fame, now at Apple I believe.) He was a class act and I learned a lot from him. I got to watch over his shoulder as he implemented a Twisted client for memcache. Woo!
Final take away: you can get away with ALL KINDS of crazy B.S. when you're sitting in a fire-hydrant of money.
(And how much would anyone reading this like to bet me that I'll never see a penny from the IPO? Please? It's the only way I'll make anything from being involved in this weird and wonderful enterprise)