I understand. You're completely right about Sidekiq. For RabbitMQ - you can have jobs persisted to disk on all nodes in the cluster which is like a backup. And

I understand. You're completely right about Sidekiq. For RabbitMQ - you can have jobs persisted to disk on all nodes in the cluster which is like a backup. And of course transactional messaging is a problem that needs special attention.