Partitioned Export with COPY Generates Excessively Many Small Files within Each Partition

[aby-kuruvilla-clear]

Duckdb Version: 1.1.3 and 1.2.0

When using the COPY command to export data into Parquet format with a partition key (e.g., batch_id), DuckDB produces multiple very small files (approximately 50–100KB each). In our use case—exporting around 100 million line items (1 crore) into batches of 5,000 line items each—this results in roughly 20,000 separate partition directories and small file fragments within each.
Current output

\batch_id=1000\
     records_0.parquet (30kb)
     records_1.parquet(70kb)

Expected output

\batch_id=1000\
     records_0.parquet (100kb)

Command used to export :

COPY my_schema.my_table 
TO '/path/to/export' 
(FORMAT 'parquet', PARTITION_BY (batch_id), OVERWRITE_OR_IGNORE, FILENAME_PATTERN 'records_{i}');