Function first_batch_at_or_after 

pub fn first_batch_at_or_after(
    data: &[u8],
    floor: LogOffset,
) -> Option<RecordBatch>

Decodes batches from data and returns the first one whose last offset is >= floor.

It skips the batches at the start of the returned byte range that the offset index pointed at but that do not cover the requested offset. Kafka offset indexes are sparse, so such batches occur.