When the message reaches our Kafka Connector that contains headers, Kafka automatically deserialises those headers. Because Kafka doesn't know about what types things were originally, things just get deserialised to the most obvious type. So that means numeric strings end up as a numbers. There’s a problem with Realtime whereby large numbers in headers breaks things.
headers shouldn’t be in extras.kafka.headers but instead simply in extras.headers - the former screws around with other things.
This PR fix two problems with Kafka headers:
extras.kafka.headersbut instead simply inextras.headers- the former screws around with other things.