ElasticsearchIndexInit jobs run by platform api seem to run occasionally twice and therefore fail because on the second run the index is already made.

The second run will always fail because the success criteria for the job are string checks that ensure the job actually made a new index and *not* that it checked/confirmed/updated an existing index.

Discovered while trying to deploy T361556

Perhaps related to the Horizon work (T342866)

Is this perhaps a race condition where two jobs run concurrently? Is it a timeout issue where job 1 succeeds but after a timeout means it is marked as failed?