I've not tried the beta version, but yes... I use \p{L} and it's variants all the time in v0.5.3. The PCRE regex engine definitely supports it.
And I have no idea what might be different, but:
Code:
(?<=\b\p{Lu})([^\s]+)(?=[^<>]*</h2>)
finds:
Code:
<h2>AARON A. AARONSON</h2>
just fine in my installation of 0.5.3 and:
and clicking "Replace All" (or stepping through one word at a time) changes it to:
Code:
<h2>Aaron A. Aaronson</h2>
It worked for all three test cases presented. I just double-checked again, so I'm not sure why it wouldn't work for you.
EDIT: Oh, and it doesn't match the WHOLE of <h2>AARON A. AARONSON</h2>, but rather it matches each capitalized word one at a time.
EDIT2: you are correct that it doesn't work in the beta--at least it doesn't with my built-from-source-on-ubuntu-64-bit version. I hope that's a simple fix (or already a known bug).