Logging levels should be revisited because:

• we always have whole logs of youtube-dl operations, even without --debug
• we always have logs of download / upload operations on S3, even without --debug
• when --debug is not set, we do not have details of the fatal exception catched at entrypoint level, meaning we have to start again the scraper while since it is fatal, we know we will probably need it anyway
• when --debug is not set, we do not have details of all preparatory work done to fetch the list of videos from topic / playlist, while it might take a long time

Probably other stuff to fine tune as well (e.g. subtitles processing?)