This is definitely another case of "you can check, you know" Overwhelming code review is absolutely a reason to reject changes, but if someone is collaborative? You can check the work. Trust isn't something that one proves. It's something we extend to others — even strangers — and evaluate the results. It’s one reason I think the 'web of trust' model is only weak information, not strong. (An echo of 'key signing parties’: It's not as if someone's presence and proving identity is the same as using their keys well and signing the right things. Yet the construction of the system of GPG is that trust is binary. This is not correct.)