· Why wrapping Kafka in HTTP push bridges destroys backpressure, how batch offset commits cause silent data loss, and why adding a second queue behind an HTTP endpoint is an architectural anti-pattern.
· How Amazon SQS bills by 64 KiB chunk, what raising MaximumMessageSize to 1 MiB costs, how the Extended Client uses S3, and when a pointer to data you already stored is cheaper than all of that.
