I don't mind them testing things, but they are normally done in beta testing.
I agree with this but, it's not really possible to test the effectiveness of an ad format on any truly representative focus group the way just putting it out there and seeing what happens can. Not saying I like being a guinea pig but I'll bet it's the fastest way to get the market research data their looking for. Testing to make sure something works consistently in all browsers vs. testing to see if it's effective from a marketing perspective are quite different.
No amount of testing will alter the 728 advert from being aligned to the left.
Frankly, the day this happened I was ecstatic. The centering of those banner ads was a peeve of mine for years (as are designers who insist on center-aligning all their text). At least now I can predict where the visual weight of the text is going to be. Before, the options for that space were wildly unpredictable, given the variety of ways the space could be broken down with one to many ads. Also, I grew up reading books and magazines with either left alignment or justified alignment and I design that way. The change was one of the most welcome things to come down the pipe for me in a long time. Having said that, one would think it wouldn't be that difficult to add ad alignment option to text ads. Right now, Smart Sizing (along with the horizontal, vertical and rectangle parameters combined with container div size restrictions) for responsive ads allows for some very interesting ad spaces to be achieved and a similar type of parameter for alignment would be a very nice improvement.
Early on, I was commenting about the pretty dramatic differences in these ads from browser to browser. I can't abide Chrome on my computer but know a few who use it and get feedback from them where testing my sites are concerned. Ads are not the only thing that renders differently from browser to browser. Drop down lists, radio buttons, heck, even the way broken image link placeholders appear on the page, vary from browser to browser. Part of this is the way the browser renders an element and part of it is plain old branding (unique button shapes, menu items, etc). Each browser has it's own feel and I know when I'm not using my favorite browser, it feels foreign in a way. It's not that surprising that G would try to tailor the ads (e.g. optimize them) to each specific browser. Just as "beauty is in the eye of the beholder," "the Internet is what our browser renders". And given the fact that everyone has a different browsing history, different habits, etc. this diversity is only going to get more extreme in the future, I believe. There may be some simple psychological reason why Chrome users interact more with one kind of button than IE users. One thing I'm fairly sure of is that G isn't doing this on a lark. Every thing they're doing is being analyzed and tracked and things are gravitating toward what works best in a great many different situations and environments. It's not a one-size-fits-all Adsense anymore and I don't think we've seen the tip of the experimentation iceberg yet. My conclusion is that this is necessary and will be good for future earnings. The tricky part here is keeping up with the changes and making adjustments when necessary to get the best advantage possible from these new features.
Added: I'll just quickly add that it's easier to complain about a feature (giant titles for example) than it is to come up with a strategy of how to work that feature into the visual language of your website. I'm finding that there are significant rewards to be had for the effort though.