Hey Jay Gorrell thanks for reporting the issue and providing the reproducibles. The Path implementation deliberately is not creating deep copies for performance and resource usage reasons. While that’s so, there’s a problem you’ve reported. It comes from the fact that you are passing a null value as a nodeName into receiver.indexedValue(..). This results in a created path missing a container element node; you can see that by inspecting the nodes in the path or by calling a Node#getIndex() either on the node or nodeImpl in your test case.

If you'd like to make it work on with the version of the validator you are using right now – you’d have to pass a nonnull node name. Keep in mind that that will change the string representation of the path (that’s if you care about how it looks).

In the meantime, I’ll submit a patch that should address the problem.