The spec recommends but does not require breadth first traversal. In a meeting we discussed the security implications of not requiring this, impact on our ability to do incremental verification of incoming blocks. The conclusion was that car mirror implies a certain amount of trust, so we are keeping this as recommended. This issue is just to remind me to explicitly call this out in the spec, not to change the spec recommendations.
The spec recommends but does not require breadth first traversal. In a meeting we discussed the security implications of not requiring this, impact on our ability to do incremental verification of incoming blocks. The conclusion was that car mirror implies a certain amount of trust, so we are keeping this as recommended. This issue is just to remind me to explicitly call this out in the spec, not to change the spec recommendations.