It's an interesting empirical question how much of what we do is sub-optimal.

Sub-optimal relative to what? To what a hypothetical God/AI with unlimited computing power would recommend? Well, we don't have access to that kind of computing power.

I read it as rejection of "tradition" as a stand-alone justification.

As Nick Szabo points out in this essay, tradition often contains wisdom that would be computationally infeasible recover from first principals. So yes, all other things being equal, you should accept "tradition" as a stand-alone justification. If all other things aren't equal, then you should treat the existence of the tradition as evidence to be incorporated like other.