Improve RabbitMQ

Goal

Correct the selected RabbitMQ publish or consumption behavior across routing, confirms, effects and acknowledgments.

Domain rules

• Make topology changes compatible and staged; avoid changing immutable queue arguments in place without migration/drain strategy.
• Use confirms plus unroutable-return handling for required delivery assurance and keep publish retry idempotent.
• Place ack/nack after the real processing boundary, bound requeue loops, and route terminal failures with complete diagnostic context.
• Tune prefetch from work cost/downstream capacity rather than using concurrency as an unbounded throughput lever.
• Own connection/channel lifecycle explicitly and re-declare/re-subscribe only through the established recovery mechanism.

Validation

• Run topology declaration and integration tests against a supported RabbitMQ version and policy profile.
• Exercise unroutable publish, confirm ambiguity, consumer crash, repeated redelivery, connection/channel recovery, blocked broker, and graceful shutdown.
• Verify queue depth/age, unacked count, duplicates, ordering scope, DLQ metadata, and recovery without destructive production testing.

Repair the maintained topology, publisher acknowledgement handling, consumer completion or connection/channel recovery owner using a fixed message/operation. Preserve immutable queue-argument compatibility and its migration strategy. Feed the candidate settings and original event order to the recovery check; repair remaining routing/effect/subscription failures before Patch Checking rather than suppressing returns or dropping unacked work.