drupflare/worker - v1.0.0
    Preparing search index...

    Variable DUMP_CHARS_PER_CHUNKConst

    DUMP_CHARS_PER_CHUNK: 1000000 = 1_000_000

    Characters per chunk.

    BYTES, and the difference from the restore is the point. IMPORT_STATEMENTS_PER_CHUNK sizes a restore chunk by statement COUNT because a replay WRITES, and rows written is the meter that binds regeneration. An export writes nothing; what it spends is the memory to hold the encoded text and the CPU to hex it, both proportional to bytes. Sized by statement count instead, 40 cfw_file_chunk rows -- 16 million characters -- would pass as one small chunk.