Improve process of flushing caches upon deployments
@rdavila and I observed that
rake cache:clear step in omnibus seemed to take a while. We also noticed other workers were running this after a deploy, which slowed things down. I wonder:
Are the number of cache keys growing larger for some reason? (A cursory glance of the Redis keys has me wondering if Sidekiq-limit-fetch is causing things to balloon.
Do we really need to run this on every worker? I would hope the blessed worker would be enough. Should we disable it for the other workers?