If the initial code had been properly optimized then starting difficulty could have been nearly 9 and we wouldn't be seeing such a short block time. This is a lesson, if your going to release a new algorithm for CPU mining, optimize the CPU code or else you've get quasi-ASIC like behavior in which hash rates explode and the coins may end up monopolized by people who have access to the high potency mining ability even if this is by software rather then hardware.
Starting difficulty would have still been where Sunny set it, 7. I don't know what the distribution looks like, but I am still finding blocks on 5 micro instances and a laptop running while I am sleeping. If difficulty adjustments took two weeks, I would agree with you, but it's dynamic.