PR gpodder#1094 generated an HTML description for any episode that lacked one. That however increased the database size (almost double in worst case) because it was storing both text and html versions ...