I emailed Sun asking about the Drag and Drop issue, and they gave the following response:
This is correct; if you return to a Drag and Drop question that you've already answered, you will need to re-answer the question. This is not a bug - due to the test driver configuration requirements, the system was specifically designed to function in this manner. While reviewing questions, if you select a Drag and Drop question for review, the system will provide a pop-up warning stating that you are about to open a Drag and Drop question, which will require re-answering the question. At this point, you able to abort the review of that item, and retain your original response to that question.
There you go, folks. It's not a bug, it's a feature. It was "specifically designed" that way.
Personally, I think that's an unconvincing answer. If software has a problem, it's still a problem even though it may have been specifically designed to behave that way. To read between the lines, I presume by the "test driver configuration requirements" they mean this is a limitation of their framework, and it would be too much effort (i.e. cost) to fix. (One wonders, of course, exactly how much effort would be needed to write testing software that works well...)
Sorry if this is turning into a rant, and it's not an enormous issue, but it catches people out. If you want to "review" a question, it would help if you could also review your answer, no?
Inquisition: open-source mock exam simulator for SCJP and SCWCD