We’re using pt-table-checksum for replica consistency checks. It relies on the STATEMENT binlog format, but Debezium’s MySQL source connector interprets these queries as DML.
Example:

REPLACE INTO internal.checksums_kafka_testing SELECT test statement based AS crc FROM user WHERE ((id >= 2611069)) AND ((id <= 2771980))

mysql connector version: 3.2.1 Final
config:
"topic.prefix": "cdc_mysql",
"schema.history.internal.kafka.topic": "topic_schema_history",
"database.server.name": "debezium_cdc_mysql",
"database.include.list": "user",
"connector.class": "io.debezium.connector.mysql.MySqlConnector",
"database.allowPublicKeyRetrieval": "true",
"transforms": "unwrap",
"transforms.unwrap.type": "io.debezium.transforms.ExtractNewRecordState",
"transforms.unwrap.add.fields": "op,table,source.ts_ms",
"transforms.unwrap.delete.tombstone.handling.mode": "rewrite-with-tombstone",
"database.hostname": "<hostname>",
"database.user": "<user_name>",
"database.password": "<password>",
"database.port": "3306",
"output.data.format": "AVRO",
"offsets.storage.topic": "_debezium_offsets",
"tasks.max": "1",
"internal.database.history.ddl.filter": "REPLACE INTO*",
"schema.history.internal.store.only.captured.databases.ddl": "true",
"schema.history.internal.skip.unparseable.ddl": "true",
"key.converter": "io.confluent.connect.avro.AvroConverter",
"value.converter": "io.confluent.connect.avro.AvroConverter",
"database.connectionTimeZone": "America/Los_Angeles",
"snapshot.mode": "recovery",
"snapshot.locking.mode": "none"