Using the Acid3 test in this context is misleading. Acid3 tests a random selection features to ensure they work corrrectly. In this case, Firefox 3.5 fails on 7 of the tests and 5 of these are due to SMIL (pretty much SVG animation) and SVG fonts. They are obscure features, and don't provide a compelling reason to choose Safari over Firefox, least of all for development where you won't be able to use those features in the real world until other browsers implement them. Whatsmore, in designing Acid3, Hixie deliberately picked on browser bugs so the test didn't favour one vendor, irrespective of how they effect the real world - taking a single number out of context and point at that to prove the innate superiority of one browser over another, or that one browser follows "better standards", is incredibly simplistic, let along drawing conclusions that sites will be easier to port because of it. At worst, all it could tell you is that one vendor is focusing on passing a test and another isn't.
I don't really see the pre-optimisation as a real-world issue, and I seriously doubt bugs creep in through it. It feels very much like somebody making a problem out of nothing.
All projects have long niggles and bugs that have been there since the dawn of time. The fact that Firefox has bugs spanning years is more of a sign that the Mozilla project has been around for years than inehrent tardiness.
I also don't really understand the conclusion that "OS X is the only system that allows me to test all major browsers side-by-side by means of virtualization software". In actuality, if testing is your priority, then Windows would probably be the platform of choice as it's the only platform that can run all the major browsers (IE6, IE7, IE8, Firefox, Safari and Chrome) natively. Linux can also virtualise all the browsers not available natively through virtualisation of Windows, no different from Mac OS X.