Function split_at_next_series 

fn split_at_next_series(
    batch: RecordBatch,
    current_primary_key: &[u8],
) -> Result<SeriesBoundarySplit>

Splits batch at the first row whose encoded primary key differs from current_primary_key, the last primary key written to the current file.

The writer finishes an oversized file at such a series boundary so that the primary key ranges of output files never overlap, which allows pickers like TWCS to detect overlapping files by their time and primary key ranges. A series is never split in the middle: if no row starts a new series, the whole batch stays in the current file even if it already exceeds the size limit.