Since 3.6 csv.DictReader emits OrderedDicts instead of regular dicts. Since regular dicts are ordered in 3.7, this is an overkill. It would be nice to add a configuration option for a return type. This is an easy issue.
For csv.DictReader and collections.namedtuple._asdict, it might be cleaner to just deprecate the OrderedDict in favor of returning regular dict. I suspect that in these contexts, no one really wants an actual OrderedDict.