Originally Posted by kovidgoyal
Do not look at the source html, that is not what the search and replace rules run on, use the wizard in calibre to look at the actual html that the rules run on.
I have. Like I said, when I test my regex the wizard finds all of the paragraphs just fine, but on conversion it misses every last one. This is true even if I use the unicode version of the non-breaking space (i.e. cut-paste from character viewer).