Split too small and you sever the context that made a passage meaningful. Split too large and the embedding averages several topics into something that matches nothing well.
Teams routinely discover that chunking strategy affects answer quality more than which embedding model or which LLM they chose.