I had huge problems on server 3 weeks after a 6 TB migration from other DB. I think it's sad to rewrite all data twice.
You could always COPY FREEZE into a new table, then move any remaining rows, and finally rename the tables. Should be a win if the majority of the table is newly-imported data. Another thing you might look at is increasing your three week forced freeze window by adjusting params and/or decreasing the number of transactions your cluster is using. (granted, still the same overall write volume). You could also do manual vacuum freezes at a time better for you (since you mention autovac is doing this at unpredictable times).
