Replying to myself to give a status update.
FYI, I'm also considering to extend the ICU regexp matcher with the bracket extension mentioned above if there is a chance that it gets into upstream ICU.
For the convenience of a smooth upgrade experience I committed r1234777 so that AOO now emulates the \< and \> expressions by mapping them to \b for matching word boundaries.
Using ICU's powerful regular expression engine has quite some benefits for conforming to the Unicode standard. E.g. finding word boundaries now respects UAX#29 (Unicode Standard Annex #29: Text Segmentation) whereas the old engine used a rather simple heuristic. As an example it couldn't find word boundaries in scripts like Thai that don't use spaces or punctuation symbols for separating words.
I'll update the release notes accordingly. Herbert
