Comment 11 for bug 1233113

The valgrind is in our code.

I am suspicious It is looking like something to do with the boost intrusive pointer, where I am finding the reference counter is not all the way back to zero when the storage engine is closed

Aside, having multiple cursor type objects gets very confusing with debugging after a while

i.e. open_query::oqgraph_cursor vs. oqgraph3::cursor classes

once we get merged into 10.0 I think I will look at refactoring some of that with a different naming convention for the 'cursor' objects that walk the graph rather than map onto the database - maybe 'walker' ??