At least if my interpretations of callgrind are correct, it's exactly the same.

You can use a typedef in this scope, makes the long lines easier to read
https://en.cppreference.com/w/cpp/language/namespace_alias

typedef TagLib::ID3v2::UserTextIdentificationFrame IdFrame;
auto trackGainFrame = IdFrame::find(mpegFile.ID3v2Tag(), "replaygain_track_gain");
if (trackGainFrame) { ...

Using a distinct name for each instance makes it clearer these are independent frames.

Remove the extra space inside the parentheses.

you can use TStringToQString
data.replayGainTrackGain = TStringToQString(userTextFrame->fieldList().back());

fieldList() may return an empty list, please add a check

toCString() was actually completely unnecessary. I removed it since convertWChartsToQString() avoids the UTF16->UTF8->UTF16 conversion of TStringToQString

This is definitely broken:
http://taglib.org/api/classTagLib_1_1String.html#a0ef8ad270d710863e0bb1c1b18cdb95d

const wchar_t* TagLib::String::toCWString () const

Returns a standard C-style (null-terminated) wide character version of this String. The returned string is encoded in UTF-16 (without BOM/CPU byte order), not UTF-32 even if wchar_t is 32-bit wide.

https://github.com/qt/qtbase/blob/c5307203f5c0b0e588cc93e70764c090dd4c2ce0/src/corelib/tools/qstring.h#L1027

inline QString QString::fromWCharArray(const wchar_t *string, int size)
{
    return sizeof(wchar_t) == sizeof(QChar) ? fromUtf16(reinterpret_cast<const ushort *>(string), size)
                                            : fromUcs4(reinterpret_cast<const uint *>(string), size);
}

Fortunately, platforms with wchar_t == wchar32_t are uncommon ...

Maybe better // Check if there are any User Text Identification Frames (TXXX) at all
Technically, the UserTextIdentificationFrame::find(...) below would be sufficient to lookup the tags.

Stray space in if (...isEmpty() )

This is used throughout the taglibextractor and should then be corrected everywhere in a different patch imho.

You're right, the idea was to skip the calls to the replaygain tags if there aren't any TXXX tags

You should check if the value can be converted to double, and probably also if it is in a sane range.

Can you do a followup-patch to fix this?

Yes, will do after this patch landed

I added the checks for a successful conversion, but in my opinion kfilemetadata should not judge if the result is within a reasonable range. This should be up to the application.