Right-size sidekiq nodes
Based on the last 7 days of metrics we're grossly over-provisioned for sidekiq server size.
|Priority||Mem %||CPU %|
We should consider reducing the instance size for these nodes, and possibly adding a few more instances for the queues that need them. We should be able to safely get to 70% utilization on our most active queues without a problem.
This would save us quite a lot of production resources. The napkin math says 50% of an engineer's salary.