Like Michael I'd love to see some data as well. Speculating from my own perspective I would imaging there is a fine line and would depend on how the animation is implemented.
I would, and I expect a lot of people would, associate a 'buzz' with a negative response. We have all sorts of negative associations here 'angry buzzing', 'electrical buzz/zap', or 'alarm buzzers'. So visually I think a rapid back and forth shaking would be seen as a buzz and generally associated negatively. Although as Michael points out it can now also be seen as vibration like in mobile phones which is less likely to have negative connotations and more likely to mean pay attention. If I just yell "Hey, look at me" without giving any additional cues I'm going to confuse you.
Not all of these animations look like a 'buzz' though. The example you gave looks like a gentle shaking back and forth and the 3D aspect makes it look more like twisting than buzzing or vibrating. It could be seen as 'shaking your head no' but it's more subtle and possibly less clear.
The addition of other visual clues, or audible clues helps the situation but, it raises an important question about the whole practice.
Should we be using visual (or audible) cues alone? I would hate to see this show up in any sort of situation where accessibility is important (ALL OF THEM). How do you communicate to the deaf or blind user that the window shook. If they've never seen a head shake or a buzzer buzz would they even know what it was if they could tell?
Look for research or more research needs to be done. Shaking probably works but should never be your only means of notification, especially when accessibility is involved.