Just to clarify that we were planning to use the elasticsearch filter plugin to retrieve previous data from Elastic for enrichment. This saves us the problem of retrieving extra data from the DB, but adds a problem of "how do I know site information has been saved and is available" when logging page views from within the same second.
This may sound like a contrived use-case, but it's just the simplest example - there are some other cases we need to do similar enrichment with a dependency on "created" events - sharing during creation is the first thing that comes to my head, and so on.
Edit
I have found a blog post - Document Processing and Elasticsearch which covers some of the options in more detail for us. There is also another discussion on this forum (Post processing of aggregation data) where somebody is having a similar problem:
We don't want to denormalize our event data to contain all of the other relevant info besides the ids [...] Anyone out there have a good solution for storing ids in elasticsearch but then being able to associate those ids with their labels (from another data store but can be made available via a rest endpoint or by syncing them into elasticsearch) for presentation purposes?
Obviously we will now be doing some reading around mappings and custom asynchronous plugins.
Ta,
-- Craig