Great article, a lot of these points really hit home with me, as we've been using queues to process a lot of our data in the last few years. I think if you cons

Great article, a lot of these points really hit home with me, as we've been using queues to process a lot of our data in the last few years. I think if you consider your queue being fine while having items backed up under normal conditions, you're going to have serious problems to deal with as soon as you hit 1) more input 2) you get a nasty bug in your system 3) often a combination of both. As an aside, I really enjoy working with AWS' SQS queues, as they allow you to define a maximum number of reads + a redrive policy. So you can throw, for example, items in a "dead messages" queue if they were processed unsuccessfully 2 times. We use this to replay data, stick it almost immediately in our unit tests, and improve software incrementally this way.