The cost of finishing the rewrite isn't that much (and, Dirk, it's really a cleaning up, not a total reimplementation, organizing pieces into logical namespaces, simplifying the datamodel slightly, etc).
The cost lies in munging all the client code to make use of the new version, and that would include third-party code not controlled by us, i.e. anyone who's customized a .LRN portlet or (as we're now seeing) uses the portal package outside .LRN.
And (much worse) the writing of upgrade scripts for current .LRN users.
As usual it's the support tail, not the development per se, that's the nasty bit and I just don't think it's worth it ...