Add a step to flush unneeded sidekiq jobs
Noted in 2018-07-24 failover attempt
As part of the failover, we want to drain all sidekiq jobs on the primary. This is to ensure that work is not lost. However, not all jobs do equally important work, and this step is time-sensitive.
We should collate a list of sidekiq queues that we can safely delete all the jobs in, and update the failover document with a command to just clear those queues, instead of waiting for them to drain.
Sidekiq jobs related to at least reactive caching, project mirroring and Pages custom domain verification, for instance, can be safely thrown away.
/cc @jarv @ahmadsherif any particular ideas aside from the above?