Each message carries the information the server needs or identifies records it can retrieve. A request does not need to return to the worker that served an earlier request.
This can simplify scaling and recovery across workers. Applications can still store durable data separately when the task requires it.
ELI5
For example, a service can accept a request to fetch an item when that request identifies the item and who is allowed to read it.
Any eligible worker can handle the request even if the previous request went to another worker.
Does stateless mean the application stores no data?
A stateless request can still read or update authorized durable records. It does not rely on remembered state in the worker between requests.
Why can this help with scaling?
Requests can be distributed across eligible workers without sending every request in one conversation back to the same worker.

