As you mentioned Remarkable file format is a PITA to extract highlights, one thing that helped a lot was to add an OCR fix phase that uses Gemini flash model to fix common OCR errors and to merge single highlights that are across pages.
It leaves aside power user features (e.g. emails, GitHub Actions to sync when laptop is asleep, etc.), which are listed here: https://distillate.dev/power-users.html