[MyHonestOpinon]

Indeed. You want to have small buffers, because big buffers increase WIP (Work In Progress), product gets damaged, problems are hidden, etc. That follows the JIT (Just In Time) ideas. But, you still want to have buffers to decouple both operations. You also want to constraint the size of the buffer, where the producer stops when the buffer is full. In the case of queues (As in AWS SQS), I am not aware of any limits where you can also stop the producer when the queue reaches a size. I believe it is desirable just from my experience on Manufacturing systems back in the day.