The recent PD4ML builds (both v3 and v4) refactor RTF output logic to improve conversion performance by bulky input HTML documents.

Now, even in extreme situations, the performance is comparable with PDF output and very often outperforms it.