diff --git a/appview/database/migrations/0006_phase2_catalog_subjects.sql b/appview/database/migrations/0006_phase2_catalog_subjects.sql
index 2424635..bab2e7d 100644
--- a/appview/database/migrations/0006_phase2_catalog_subjects.sql
+++ b/appview/database/migrations/0006_phase2_catalog_subjects.sql
@@ -2,8 +2,22 @@
 -- introduce the episode/podcast catalogs. Every subject-bearing collection
 -- (comment, recommendation, bookmark, episode_state, subscription) now
 -- references a catalog record via {subject_uri, subject_cid}.
+
+-- ----- Pre-Phase-2 wipe -------------------------------------------------
+-- Legacy rows in the social tables reference Podcast Index integer IDs that
+-- the Phase 2 schema no longer carries on the record. There is no automatic
+-- way to project those rows onto catalog strongRefs — the catalog itself is
+-- empty at this point, and even a hydrated catalog wouldn't recover rows
+-- whose PI IDs Podcast Index has since reissued.
 --
--- Nothing is in production, so we drop legacy columns without a backfill.
+-- Per the project's "nothing in production yet, wipe freely" stance, we
+-- truncate before altering. Authoritative data lives in users' PDS repos;
+-- the firehose + xyz.effem.admin.backfillUser rebuild these tables.
+TRUNCATE TABLE "comments";
+TRUNCATE TABLE "recommendations";
+TRUNCATE TABLE "bookmarks";
+TRUNCATE TABLE "episode_states";
+TRUNCATE TABLE "subscriptions";
 
 -- ----- Catalog tables ----------------------------------------------------