Recently the collect method in the run blundler was refactored to reduce mixing of code paths. Previously _maybe_format_datakeys_with_stream_name would take one of two different data structures layouts (a new and an older) and would format it respectively to it's type. Now these two paths have been separated so that _maybe_format_datakeys_with_stream_name is called with one structure type in declare_stream, and with a different in _describe_collect. Instead of having one method for both it would be worth doing these seperately.