Those early days are over and 3.x is intentionally designed to be more rational and consistent.
The issue being that is *always* the case. In the early python 2 days, they thought the 'early' days were over. I haven't dealt with python 3 with sufficient depth to be keenly aware of any real gotchas, but the fact they decided to add back in the explicit unicode syntax is a sign that they have at least continued to indulge in flux to fix bad design decisions. In that specific case, I don't see a downside for the increased ability to have python2/3 agnostic code so I won't declare any example of breaking 3.x series code with that. It seems clear to me that the python language can't quite exercise enough restraint in their enthusiasm for their vision of improved syntax and features to walk away confident that code won't break in a couple of 3.x generations.
It's almost like a curse, the more popular and energetic a language implementation is, the more likely it is to experience some incompatible evolution.