1. I wouldn’t emphasise as much the software paradigm, and would probably scrap the part “What if software were developed through closed models”. I don’t think it brings a lot to the discussion of publication, except to show that open processes are successful in programming.

2. The speed of innovation is not as clear cut… True, opensource softwares have been at the hear of important innovations. But commercial companies also have been very quick to foster major developments… They also have more focus on usability/GUI than open source contributors might do.

3. I am also at the LSE (Economics) and have certainly been given the opposite advice: to be bold and daring! So no generalisation… it’s probably highly dependant on the topic, and its internal dynamism (the two issues being obviously linked in both directions).

4. Also, I think one point that you do not mention at all is the potential unintended consequences of non-anonymous refereeing.

In a given specialty, there are not that many people worldwide that can comment. It means that people working on the same topic as you are going to be people you interact repeatedly within your career.
If you are not anonymous anymore, and you get rejected for (what you believe to be) the wrong reason, you’d probably be mad at the referee and will want to reject his next work… In the long term, it’s probably for the quality of the research if it’s very hard to publish and everyone tries to shoot down bad ideas, but one also needs to think about the incentives on the writer’s side… If it’s so hard, maybe it’s not that fun as a career.
The argumetn goes also the way around: if you have to ref someone who is known to be influential (not necess. for publication but for other academic issues), you might not want to reject that person if you are going to apply to his university for instance.

Just to say that these “reputation” issue should be taken into account when considering non-anonymous refereeing, which is part of the Open-Process you propose.