I have a setup with celery and a redis broker, messages are being handled by two workers. We send alot of task at once (~20k), most of which being retried due to an api limit (10/minutes). That means alot of tasks are being retried many many times. Each retries having a delay less than a minute.
The issue I face is that, I see tasks being tagged as retry in flower and nothing happens during ~1h, after that, all tasks are being treated by the workers and then, again, I must wait 1 hours before seeing tasks sent to workers.
I have no idea where to look for and I would gladly welcome any suggestion.