Without proper grapheme support in the regular expression library, it is impossible to correctly process Unicode.  And the very least, one needs the \X escape supported, which is an extended grapheme cluster per UTS#18. This escape is supported by many regex libraries, include Perl's own and of course PCRE (and thence PHP, the standard ICU library, and Matthew Barnett's replacement regex library for Python.

How do you process a string by graphemes if you cannot split on \X?  How can you avoid splitting a grapheme into silly pieces if you cannot match one?  How do I match the letter O no matter what diacritics have been applied to it otherwise?  A match of (?=O)\X against an NFD string is by far the simplest and best way.

This is necessary for a wide variety of reasons.  Adding \pM and \PM go a little ways, but not far enough, because that is not how grapheme clusters are defined.  You need a proper \X.

As I said on #12734 and #12731, if the 'regex' module address this issue, we should just wait until we include it in the stdlib.

Again, I would be disappointed if the re (_sre) module could not be fixed.  It is a reasonable feature request.