I'm seeing an interesting case, when multiple queue consumers causes failures in a downstream systems a.k.a.thundering herd effect.
Many jobs fail, they all get enqueued to try again in a static interval and bring down the downstream system yet again. This repeats until the retry limit is exhausted.
Introduced delay randomness prevents thundering herd effect. The randomness can be controlled via$jitter parameter.

Open question: I added$jitter parameter with a default value of0.1 (as inhttp client's generic retry), but maybe it should be set as0 to keep code in a BC fashion?