Looks as though the problems with the scanner were being precipitated by having too many simultaneous postgresql queries. Introducing more randomness into the scanner has let it run for a few hours at a time so far. Have begun producing a trivial summary/overview in simple text which makes it possible to confirm that the scanner is behaving as expected.

Now to convert that to a graphical layout and then figure out what to do about storage/access (I think I'll generate from the Twisted scanner into a file for access via Zope to keep things simple).