That's an excellent example. Mostly because I've probably had to replace simple retry loops more than once after they triggered various kinds of death spirals.

Even recently, an engineer on my team spotted a retry technique, where the system tried to check if the DB's replication mechanism was slowing down, then would just re-enqueue the message for later. This ended up doing a bonkers amount of wasted effort because when DB replication started getting overwhelmed, messages would just constantly be re-enqueued and not ever processed. So the queue system ended up just having a bonkers amount of messages that were just popping off the queue, doing nothing, and hopping right back on.