After Monday's overnight test run of the merge code, we've made a bunch of tweaks and we're now re-running that code so we can get a cleaner output. We're only doing stage 1 (processing each LAC record and merging a CGWP record if there's a candidate); I've written a second process to check for a) cases where the same CGWP record has been merged with more than one LAC record, and b) cases where CGWP records have not been merged with any LAC record; I'm testing that with a tweaked version of the old output, but eventually that will be run against the merged data we're currently producing to provide a possible list of actionable issues for a human to examine.