[Bug]: Current offset is higher than last offset

[vector-mj]

Description

After a network incident(without any PVC issue or data loss at disk level) in our data center, we see that the current offset(consumer group offset) of some partitions in some topics is higher than the last offset(last record offset in the partition) of that partition.
I read relevant Kafka documents about this issue, then I noticed that this issue never should happen because when you try to fetch or commit an offset that doesn't exist(higher or lower than Kafka offset range), Kafka returns an OffsetOutOfRangeException
also when there is no committed position in consumer group Kafka will use auto.offset.reset configuration.

Question:

• Can this issue be from Sarama?

Versions

Sarama	Kafka	Go
v1.41.3	v3.2.0	1.21

Configuration

// Consumer
config.Consumer.Offsets.Initial = sarama.OffsetOldest
config.Consumer.Group.Rebalance.Strategy = sarama.BalanceStrategySticky
config.Consumer.Group.Rebalance.Timeout = 2 * time.Minute
config.Consumer.Group.Rebalance.Retry.Max = 100
config.Consumer.Return.Errors = true
config.Consumer.MaxWaitTime = time.Second

// Producer
config.Producer.RequiredAcks = sarama.WaitForLocal
config.Producer.Retry.Max = 10
config.Producer.Partitioner = sarama.NewManualPartitioner
config.Producer.Return.Successes = false
config.Producer.Compression = sarama.CompressionGZIP
config.Producer.Return.Errors = true
config.Producer.Flush.Frequency = time.Millisecond * 10

Additional Context

related issue: #1734