It's not a huge restriction since GAE will happily spin up dozens of instances without any intervention if your load requires it. Getting multiple concurrent requests into each instance will make things like local memory caching somewhat more useful though.