Currently the non-default BSON validation checks are roughly grouped by cost. This approach makes the collection validation output to be noisy with trivial issues like deprecated BSON types. Recategorizing them by severity will help because more serious errors are expected to be less common.