This is essentially the same approach as https://github.com/snowplow/snowplow/wiki/Setting-up-the-Clo....

What's neat and novel to me about this implementation is the use of delimiters vastly simplifies loading the raw S3 logs into Redshift.