Every claim that's been made in this thread is based on anecdotal experience. Maybe in 50 years software will be mature enough that we'll have actual empirical

Every claim that's been made in this thread is based on anecdotal experience. Maybe in 50 years software will be mature enough that we'll have actual empirical data to back up our design/tooling decisions. But for now, almost no one is seriously interested in testing this stuff with well designed (and well funded) studies.