I'm strongly against making the act of quality control the responsibility of the witnesses. *Evaluating* quality control is the witnesses' responsibility, but we need a new structure of rewarding testing and documentation, because bad things happen if you load that onto the witnesses.

Particularly, requiring a witness to be a code reviewer ends up requiring code reviewers to be witnesses, and popular ones, because code review is only rewarded via witness rewards. What this ends up meaning is that you lock out people who are interested in development but not in system administration and witness politics, which is the large majority of good devs. It also essentially restricts the potential working dev population to a couple dozen at a time, and makes it so that someone who has put a lot of work into understanding the Steem code can have that work devalued by a single significant witness unvote.

The people who have/want to develop witness skills should be the witnesses, while the people who have/want to develop code review skills should have a different path to success, one that suits that skillset and encourages them to communicate their work to the witnesses. We already have a strong model for how this could work in Utopian. That could be copied for Steem development if the SP for it were available.