--server-queue-requests

Holds requests that exceed --server-max-concurrency in a queue of this many slots instead of returning 429. A request is enqueued while a slot is free and runs when a render finishes; once the queue is also full, further requests get 429. The value must be at least 1.

When set without an explicit --server-max-concurrency, the concurrency cap defaults to 2.

This is an Enterprise-tier flag.