[DiapDealer]

You're under no obligation to let calibre replace your named/numbered entities with unicode characters (visible or invisible). Just turn off the editor preference that changes entities to characters immediately after typing them and remember not to use the "Fix HTML" or "Beautify" tools. Concessions always need to be made when you want a program to automatically correct/tidy some code (and I'm not suggesting there's anything wrong with needing to do so) for you. But I rarely have broken or unreadable (x)html that I open in an ebook editor anyway.

The ability to open ... edit ... save ... without the editor making any behind the scenes (and silent) changes whatsoever (yes, I know I have to be sure to turn off the metadata update when opened inside calibre) is one of this editors greatest strengths, I think.

Also, note that even if I do allow calibre to replace my entities with unicode characters, I still clearly see a yellow background around characters (in the edit/code window) that are considered "special." Sticks out like a sore thumb to my eyes (but I can understand the frustration if some characters lack this distinction). More so than the original entity did, in fact. If it's not clear to me what it is at a glance, I can "zoom" further with the cursor.

The narrow non-breaking space, the thin-space, the zero-width joiner, the soft-hyphen and other such characters/entities have played havoc with editors/devices for a long time--many don't support some or all of them at all. I'm sure if there were a simple solution, it would be implemented already.