Except most times people don't implement the surrounding infra correctly. Let's say you consume a job, do something, but that something fails. You have to set

Except most times people don't implement the surrounding infra correctly. Let's say you consume a job, do something, but that something fails. You have to set it to be rescheduled somewhere. Lets say your things writing jobs to the queue fails. That needs to be handled as well. "They will be processed later [if I use a queue]" is not true by default (without work and thought) and assuming so will get you in trouble, and will also make certain people on your team who do understand that isn't the case upset.