Sorry for not posting for such a long time here - didn't even read all the (negative) comments on previous pages.
The good news is that I finally found the time to set the machine up that I got from JDB. I can reproduce the problem, and we've started to investigate it: It doesn't seem to be a hardware problem, as all access methods work fine, and a test routine specifically written to fix this issue doesn't show any errors.
Next step is to continue searching in the software. With the hardware-test being completed without any issues found, the good news for *everyone* is that nothing has to be shipped back to Germany. This frees the logic analyzer-space on my desk, and I can move on to another project where I need that equipment. The A4000D will move to another desk where I can run test-software that I get through the net.