Think of all the awesome tools people could write with search data :)
Major things to figure out:
# replication strategy (including skipping private wikis)
# making sure we've got the resources in labs to take a replica of the prod indices
* Provides access to the community to query arbitrary data out of elasticsearch. The mediawiki search api pales in comparison to what can be done with the ES api directly.
* The Elasticsearch query and document format is, for many tasks, much easier to use. In the mysql labs database getting a page and all its information is a complicated join. In ES it is a simple query, and by default the returned document contains everything we know.
* Allows tools to be built by the community to take advantage of all this data in elasticsearch.
* Gives discovery department access to constantly updated indices in labs for analysis and research into potential changes