Unlike the other IDs which are stored in the DB, this isn't relevant
for the Celery task as it invokes a link test. Moving it into the
proxy client will also enable us to generate a second ID in the next
commits, where we start doing a link test for the failover lambda.