The general idea is that you have a chance to parallelize waiting for review with you writing more code, leading to high-velocity small PRs and never being blocked.