>>>>Fernandos explanation of log placements is spot on. A better manual test would be to call the troubling form up a dozen times.
>>>
>>>I now have one in the menu call right before the do form and another one right after final read events. I'll do more tests later today, it looked like indeed there was a jump in the memory every time this form was invoked.
>>
>>If so, add a call to the log in the errorhandler as well (handled errors can result in leaks/fragmented memory).
>
>If there would be errors, I would get them - there are no errors in that form.

You know the code base - not even triggered try-catch ? If that is not the culprit, there were some tools to log memory - probably Chrithoph either has it or a mention ;-)