@rowan, indeed, there are 2 ways of running Lambda architecture. You can find more details in this Discourse post: How to setup a Lambda architecture for Snowplow.
In short, you need to run S3 Loader on stream enriched data and EmrEtlRunner in Stream Enrich mode. This way the data will be enriched once (in Stream Enrich component) and EmrEtlRunner will be used to shred and load the data to Redshift.