[QUESTION] Using AIOKafkaProducer.start() and stop() with kerberised Kafka

What is the preferred way of using AIOKafkaProducer.start() and stop() with long-running application and with environment when you have to periodically renew kerberos ticket for accessing brocker?
If I use krbcontext for example should I include start() and stop() in same context with AIOKafkaProducer.send() or start() and stop() should be called just once in lifetime of application and only AIOKafkaProducer.send() should be called in krbcontext's `with` statement?