By default, the MongoDB Kafka source connector publishes change event data to a Kafka topic with the same name as the MongoDB namespace from which the change events originated. 

As naming restrictions for database/collection names in MongoDB allow characters that are restricted for Kafka topics (e.g whitespace), such namespaces will fail in the connector.

While topic.namespace.map can be used as a workaround, it won't be of any help for newly created databases/collections and has to be constantly updated upon failures.

The property topic.mapper looks relevant but lacks any documentation and seems complicated to implement.

IMO, a topic name sanitization should be implemented, something similar to how it's done in Debezium:  https://github.com/debezium/debezium/blob/4d577655968d2e00edb67a7e53702b8c38a17023/debezium-api/src/main/java/io/debezium/spi/topic/TopicNamingStrategy.java#L46

(Debezium MongoDB Source Connector cleans up any illegal characters for topic name by converting them to underscores)