Spring Kafka Consumer skips some offsets

We have a very simple Kafka Consumer (v 2.6.2). It is the only consumer in a consumer-group, and this group is the only one reading from a topic (with 6 partitions, and some 3 million events in it). Broker is also version 2.6.x

As we need to fulfil a "excatly-once" scenario, we took a deep look, if we really consume every event exactly once that is written to the topic. Unfortunately we found out that: The consumer sometimes skips an offset, and sometimes even a bunch of offsets of a partition.

The Consumer does not do much more than logging. An Error-Handler is configured with logging.

@KafkaListener(id = "${spring.kafka.consumer.group-id}", topics = "${topicname}")
public void onMessage(ConsumerRecord<String, BlaBlaEvent> record) throws Exception {
  // mostly logging
}

And it logs:

consuming Offset=30 Partition=2
consuming Offset=31 Partition=2
consuming Offset=32 Partition=2
consuming Offset=67 Partition=2

(did you see the Offset=67?)

Here's our application.properties

spring.kafka.bootstrap-servers=${KAFKA_BOOTSTRAP_SERVERS_URL}
spring.kafka.properties.security.protocol: SSL
spring.kafka.properties.schema.registry.url=${SCHEMA_REGISTRY_URL}
spring.kafka.properties.specific.avro.reader=true

spring.kafka.ssl.trust-store-type=PKCS12
spring.kafka.ssl.key-store-type=PKCS12
spring.kafka.ssl.key-store-password=${KEYSTORE_PASSWORD}
spring.kafka.ssl.trust-store-password=${TRUSTSTORE_PASSWORD}
spring.kafka.ssl.trust-store-location=${TRUSTSTORE_PATH}
spring.kafka.ssl.key-store-location=${KEYSTORE_PATH}

spring.kafka.consumer.client-id=BlablaNotificationReader
spring.kafka.consumer.group-id=blabla-group
spring.kafka.consumer.auto-offset-reset=none
spring.kafka.consumer.max-poll-records=1
spring.kafka.consumer.key-deserializer=org.apache.kafka.common.serialization.IntegerDeserializer
spring.kafka.consumer.value-deserializer=io.confluent.kafka.serializers.KafkaAvroDeserializer

logging.level.org.apache.kafka=DEBUG

Sidenotes:

• Consumer runs in OpenShift v3.11 (Kubernetes 1.11). No pod restart or anything happened at that time.
• No errors in the log on broker-side, nor on consumer-side. just silent skipping.
• It happens about once a week

Does anyone have an idea how to find the root of this problem?