Interesting that the 1G machine seems to run OK.

Just guessing based on other experience - but I suspect the problem is that the smaller RAM is not big enough to hold the amount of frames generated during video capture without splitting the images somehow. That would lead to lots of paging to virtual memory - which in turn aggrivates the Java memory leak problem - which leads to even more paging, etc., etc.

Note that one change in 2.2 on Windows machines was to move away from the earlier graphics approach. That might be the reason that the program size increased so much. (I agree - 70+ M seems like a lot.)