...Basically it's the same as with most of the bugs: there's no QA at Sparx.
Yip, agreed, over and over. the lack of internal testing BEFORE release is obvious by damaged
expectations concerning Feature X: always delivered half-tested and half-implemented. EA scrollbars are a perfect example.
It is understandable that software has bugs; that is the nature of humans writing code. However, methinks Sparx is lumbered with a codebase as ancient as COBOL and a team too afraid to communicate with it. This explains why "fixes" in each release always address only one
specific scenario, and disregard others. The team is either focused on more "high priority" money-generating concerns, OR every issue, bug and feature is low priority and "not serious", OR the developers just plain and simply cannot address the fix using an ancient programming tongue, like ADA.
Take, for example, this quote
Now that Tags are stereotype properties, it is self-inconsistent to NOT copy tags of the connectors when the [X] Copy Connectors checkbox is marked.
It's a beautiful example of how (clearly) no testing was done when implementing the use case for "Make Tags stereotyped properties". Otherwise, why should a paying customer need to state the
obvious?
I very much want to keep a positive attitude in life and work 
There are a lot of things much more important than EA
I still believe EA is the best (least bad?) UML tool out there currently.
Geert, you set a good example by remaining positive. Though, if you were not an EA consultant, would it be as easy to overlook annoyances of the "least bad" UML modelling tool :-)?