The Last-Event-ID and EventSource id field will be used to handle auto-resume during client disconnects. By using EventSource to connect to a KafkaSSE endpoint, your client will automatically resume from where it left off if it gets disconnected from the server. Every message sent to the client will have an id field that will be a JSON array of objects describing each latest topic, partition and either offset or timestamp seen by this client. On a reconnect, this object will be sent back as the Last-Event-ID header, and will be used by KafkaSSE to assign a KafkaConsumer to start from those offsets. If the underlying Kafka cluster supports timestamp based consumption, the Last-Event-ID header may initially provide assignments with timestamp entries instead of offset. If timestamps are provided, then Kafka will be quried for the offsets that most closely belong to the provided timestamps. This allows for historical consumption from Kafka without having to know the logical Kafka partition offsets for particular times in the past.