The only time I criticize a group's karaoke is if it's so concerned with being flashy and impressive that it forgets that people might actually want to read the content. Or if the songs aren't translated whatsoever.
At minimum, the English translation is necessary -- otherwise you've got a release that's leaving 3 minutes of content untranslated. JAlmost as bad as leaving 3 minutes of dialogue untranslated, in my view. I don't care how corny, generic, or irrelevant to the show's content a song is. If the Japanese viewers understood it, so should we.
English + no-effect romaji is fine.
English + simple k-timed romaji is good, and actually practical for singing. Effects that make syllables appear/disappear as they're sung = stupid.
Kanji - not necessary, as most viewers can't read it.
I know I'm in the minority here, but I prefer hardsub karaoke. It allows for cooler effects, which I don't mind as long as they're not too over-the-top. Plus, it means less CPU strain on my old/crappy system -- once again, minority, I know.
Simple softsubbed karaoke is fine, but I'd rather see more intensive effects like color-transformations and moves hardsubbed.
Insert karaoke is a different matter. Anything more than English + simple k-timed romaji is an unnecessary distraction. Granted, it can look cool, but some groups take things waaaay too far. Behold, 8 lines of subtitles!
And that's all hardsubbed, btw. Dialogue, lyrics, romaji, kanji, everything.
The one other facet of karaoke I take issue with is that fansubbers don't think that sane standards of readability apply to karaoke subs. Most respectable groups wouldn't use decorative cursive fonts, tiny/transparent serif fonts, dark fill + light outline, or borderless subs colored exactly like the background for main dialogue subs. (Wait, scratch that last point, they do.) But apparently it's okay to commit these eye-crimes with karaoke
