- Make new camus eventbus jumbo job: https://gerrit.wikimedia.org/r/#/c/419493/
- Let both camus jobs run and consume a few runs from both clusters.
- Disable analytics camu eventbus job.
This will cause a some duplicates to be present during the time that both camus jobs are running, but should not cause any loss as long as both jumbo and analytics replcas are relatively caught up.