A middle ground wasn't advocated, he was basically saying "let's keep the pace of innovation constant, but improve quality". Putting incremental effort into quality and testing probably requires you slow down your rate of experimentation, no?
Of course, it's an example where there's all kinds of potential for problems that could take a serious amount of time to work out. Can't keep innovation pace high with something like that plus ensure quality. So, the trick is usually to leverage as much proven stuff in one's innovation as possible. Easy to do with software at least.