Responding to your edit about "a SF startup could've built out an amazing system at a fraction of the price"-- I hear that kind of attitude a lot, I'm not so su

Responding to your edit about "a SF startup could've built out an amazing system at a fraction of the price"-- I hear that kind of attitude a lot, I'm not so sure it holds to scrutiny. Tackling the problem of a system that must be 100% reliable, synchronous, with complete uptime from day 1 is very very difficult, and not many startup environments will prepare you for it. A lot of systems that are scaling decently now (twitter was used as an example early on) had extreme early growing pains. In fact, I think it is impossible to build such a system without significant upfront cost, because a lot of the cost should go into the testing phase. It's almost as though you'd want to create a testing call center that receives realistic calls and needs to route them correctly, and have it running for months while the software is iterated on. A lot of effort should go into making the testing center as realistic as possible, and a lot of time should go into teasing out the inevitable bugs.