The whole problem rests in the fact that CKEditor (and most WYSIWYG editors) convert non Latin characters to HTML Entities...
If you watch carefully you'll see that even after applying the solutions above, they only thing that happens is to "convert" those entities without any errors, thus people see all the characters as they were supposed to be seen.
BUT underneath all text is being stored as HTML Entities and not as UTF on the Database, which is not ideal for many reasons (thing about a massive forum and the overhead of using HTML Entities for each character!!!).
The PROPER solution to this is to tell CKEditor [b]NOT[/b] to convert Greek, Cyrillic etc. characters to HTML Entities but treat them as normal UTF text. This way there is no bug at all.
The current (default) setup [b][u]does[/u][/b] make sense for users that use [b]Latin characters[/b] and very rarely have a couple of Greek characters in their posts. This way they can use a different encoding (e.g. ISO-8859) and still have those characters displayed properly (since they are stored as HTML entities).
If a board uses Greek or Cyrillic as the [b]main language[/b] then THERE IS A PROBLEM! Obviously we do not want ALL our text to be stored as HTML Entities. This is the main language so what's the point doing that?
Go to [b]public/js/3rd_party/ckeditor[/b] and add the following line to [b]ips_config.js[/b]
[CODE]config.entities_greek = false;[/CODE]
additionally, there is an option to disable entities globally but needs some testing.
[b]PROPER SOLUTION (maybe via patch or 3.2.4): [/b]
Switch this option by having a drop down box somewhere under ACP for the languages that have this issue.
[b]EVEN BETTER SOLUTION:[/b]
Combine the proper solution above with the option to use the standard editor we all loved those years.