Hi Jaceb,
 
> > I think it's just a marker, so gecko knows that this node is included
> > in collection business.
>
> Yes, it is a special case, see [1]. It's used to cannonicalize interface
> pointer so that it may be used as an identifier in cycle collector
> graph. It's not a real interface and it should not cause reference change.
>
> Alistair, your bug is probably somewhere else. If your patch "fixed" it,
> it indeed suggests a refcount-related problem somewhere else.

Yes, the patch appeared to "fix" the issue. I'll see if I can track the problem
down.

Best Regards
 Alistair Leslie-Hughes