Rolf Mertig,
I'm not sure if this is a matter of more beta testing, or of keeping a lower amount of bugs alive. There are too many reports, acknowledge as bugs, that are live for several releases now, and that are not related to some hard or still not known to humankind algorithmic capacity.
Some of the "Mathematica gets this Integration wrong" kind of bugs, for obvious reasons, can last for a long time (I said some...); but bugs related to such a fundamental data layout function, like Grid (you know that this weekend I was fixated with Grid...), that last for several releases, are, in my opinion, hard to accept.
Since we don't have an open bug list (which I can perfectly understand, or at least, a fully opened one), its difficult to say if these bugs were known prior to their first release. But even if not, we know that they were known prior to their second, third and fourth release, which makes me think that even if the Manipuate bug of the OP was known, it seems that it could perfectly have been approved for release (I know that it's a harsh thinking, and an unpleasant saying, but it's solely based on shown behaviour).
I really hope that this situation gets improved, or I will start attending the conferences with a T-shirt saying, "please get [put some functionality name here] fixed" (if WR would let me in with it :-)