I can't yet put my hands on the two earlier support requests that addressed the issue; but our investigations pointed to the problem as having emerged with the recent change made to the Punquin template to add the link in the lower left table for skipping back to previous entries. One of our volunteers did find that yes, as had long been quite stable before, if one were to remove that new link from the code, then the bleeding problem disappeared. Since we'd like to have that skip-link there, that hardly seems like it ought be the end of it. On further tinkering, I found that for the table that was the direct parent of that new link, if one changed border=0 to border=1, then the problem is also solved: the bleeding disappears. And since we don't really want a border there, such a patch could be easily enough handled by using a border color consistent with the style's color scheme, so the effect would be invisible to the users. Even so, personally it left me dumbfounded why (1) simply adding new text to a previously working table should have this bleeding effect, (2) why the effect was also reflected in the upper left corner, and (3) why simply adding a border to this one table should eliminate the problem. I saw nothing with the table structure of the Punquin page that ought to have had Navigator choking, although I've not yet run it through a rigorous validator check.
This problem has nothing to do with any override that the user might be using; and as far as I've been able to work out, it's not something that can be patched via any override. For any paid members, the problem can be fixed if the user creates a clone of Punquin and directly attends to the code in the lower lefthand corner. Of course, one solution would be for the user to upgrade to a newer and better version of Navigator (version 4 is notorious for choking on even the most basic valid HTML/CSS); and although to the best of my knowledge there's no formal, official policy on this, I doubt that one of LJ's priorities is ever going to be to revise any of the coding to have backward compatibility to Nav 4.x.
But unless someone else has any other ideas, expect that this will not be the last we hear of this one.