Currently connector retry logic relies only on communication problems, e.g. TLS error. If the endpoint is busy or rate limited there is no way to tell Zabbix server to retry after some time.

Connectors will benefit from checking HTTP response codes and retrying the operation after a delay if it's negative.