>>jedber+(OP)
Yes, I don't know any distributed task queue that stays online or backs up tasks when the broker dies. Maybe one could be built on CockroachDB? My implementation is designed to process tasks fast and if the Redis server dies (hasn't happened in 10 years) you just boot up a new one. I try to design my tasks to expect some failures and be able to re-run safely.
>>welder+d2
RabbitMQ and SQS both have redundancy built in.
There's nothing wrong with the trade off you made, but when I read distributed I assumed that you meant the task broker was distributed and therefore redundant, not the consumers.