I've repeatedly seen it broken in Opera across platforms pretty much from the first time I encountered it -- though it it's consistent in cropping up. It often becomes like people who declare %/em font sizes without redeclaring the line-heights, or using the named sizes, and the net result is all the lines overlapping.
There was an example of the behavior (from a similar cause) I pointed out in a recent thread in regard to Neilsen's websites. My desktop and laptop are both high res (1920x1200 and 1680x1050 respectively), both run large fonts/120 dpi, both running the same versions of Opera... but when I visit their site it's fine on my desktop, but my laptop gives me this:
Back when M$ first announced their new "artist formerly known as frontpage" rubbish, the website for it at that time used the no-metric line-height along with fixed-height containers (the latter being REALLY stupid), resulting in this on all my systems:
... and I've been seeing sites broken like that more and more as people omit their line-height when they change the font size -- it's why I have the rule 'if you change the font-size, change the line height' -- at which point it ends up just as many characters in most cases as just redeclaring the entire shorthand FONT property -- which is why I just do that.
It's one of those stupid little things people do to allegedly 'save time' or 'save a few bytes', where to be frank much like whitespace stripping its typically a band-aid on a knife wound, you might hide that it's there but it doesn't fix a deeper rooted problem... or the comments between sibling elements bugs which is SO easy to rectify -- put the comment after a open tag or before a closing tag, not the other way around... I repeatedly have seen people diving for vague hacks and bloated CSS when the only thing wrong was doing something like </div><!-- end section -->
Leaving off the unit is unreliable and unpredictable cross browser, at MOST it's two extra bytes to NEVER have it EVER cause a problem... That there are people even arguing in favor of something so silly is almost as mind-numbingly astounding as the people who actually defend the use of HTML 5.