tables again. In this case the traversal was not properly pushing
information pertaining to hash tables onto the stack when the
traversal occurred via encountering a hash node during traversal
of the unseen nodes of the trie. Another case, where information was
pushed onto the stack when initialziing information from a segment of
choice points was in fact working properly (a prior bug).
This fix solves a CMU-SEI problem, but I need to look at it with fresh
eyes to see if its general or not.
