@ElMiko - ahhh, that is an interesting one. And I know exactly why that would be happening too. Behind the scenes when you apply one of those fancy \1 substitution "clips", it turns the selected text into a regular expression. The PCRE engine cannot cope with the unicode line feeds which is why your first problem occurred. However it also will of course get itself confused when being passed text that contains text like your example, as it must make sure to escape the text first in case there are any other regular expression characters in it.
I will sort it out for the next build, that is a very subtle and nice bug to find, well done
.