Background workers and open connections
Return after each batch so idle connections do not keep consuming compute.
How work starts
SproutOS invokes your application when a queue has work. Handle the queue.drain event, process the supplied batch, and return. The same deployed function can serve HTTP and workflow invocations.
Return when work is done
Compute is billed in GB-seconds until the handler returns. Close database connections and do not leave a Redis subscribe, blocking read, timer, or worker loop alive. SproutOS invokes the function again when more work arrives.
Use a workflow step for work that must continue later. Split work that cannot finish within one invocation, enqueue the remainder, and return.