I’m trying to set up a RabbitMQ consumer using the amqp library. I want to process multiple messages concurrently and properly handle errors (properly reject message if a processing error occurs). How would I go about doing this?

The “Setup a consumer GenServer” example defines a single GenServer that spawns a new process for each incoming message which paired with prefetch_count > 1 allows for concurrent processing of the messages.

How do I make sure that those processes are cleaned up properly? Is spawn_link enough to shut them down when the parent GenServer is shutting down?
Should those processes live under a supervision tree?
How would I implement a timeout for processing of a message (kill it after 5seconds)?
The example says “You might also want to catch :exit signal in production code.” Why? When would that happen? How do I do that?
Are these consumer processes a use case for Tasks or DynamicSupervisor?
Or should this be set up totally differently, i.e. have a Supervisor with N workers where each one is a consumer on a shared connection?