I've dug into this one a little further and it appears that there are 6 very small shapes that are being painted which is causing the JVM to crash, or at least on windows.

I need to take a little more time to try and figure out what is so special about these shapes.

We've had a very similar test case in the past but it was too complex to narrow down the exact cause. Adding it as a second test case.

For the first example there shape counts to filter out the offending shapes are:

count++;
if (count == 562 || count == 602 || count == 606 || count == 656|| count == 750|| count == 796)

Adding this to the ShapeDrawCmd can get the PDF to render. Further investigation is needed.

Similar dashed line issue.

Similar dash issue.

Finally managed to isolate the draw operation that was causing the JVM to crash. The issue is related to a dash array values that are < 0.5 user unites. From what I can tell some encoders are using the wrong space when writing out the value or values. I've put in a little logic that converts a really small value back to users units. These avoids the dreaded jvm failure but also get the dash stroke to paint correctly.