Let the server make retry decisions, when applicable.
In some cases the server knows how to handle retry behavior better than the client. For example, if a server-side resource is exhausted and the server knows when it will come back online, it can tell the client exactly when to retry.
If this option is enabled, any supplied retry, backoff, or idempotency policies may be overridden by a recommendation from the server.
For example, the server may know it is safe to retry a non-idempotent request, or safe to retry a status code that is typically a permanent error.
