diff --git a/package.json b/package.json
index d19f62d14..a051effde 100644
--- a/package.json
+++ b/package.json
@@ -29,7 +29,8 @@
     "wait-until-postgres-is-ready:pg_isready": "until pg_isready -h localhost -p ${NEXT_PUBLIC_HEXCLAVE_PORT_PREFIX:-81}28 && pg_isready -h localhost -p ${NEXT_PUBLIC_HEXCLAVE_PORT_PREFIX:-81}34; do sleep 1; done",
     "wait-until-postgres-is-ready": "command -v pg_isready >/dev/null 2>&1 && pnpm run wait-until-postgres-is-ready:pg_isready || sleep 10",
     "wait-until-clickhouse-is-ready": "echo 'Waiting for ClickHouse to be ready...' && pnpm exec wait-on -t 10000 -v http-get://localhost:${NEXT_PUBLIC_HEXCLAVE_PORT_PREFIX:-81}36/ping || echo 'ClickHouse is not ready after 10 seconds (this is probably a problem with our wait-on setup TODO fix), continuing regardless...'",
-    "start-deps:no-delay": "pnpm pre && pnpm run deps-compose up --detach --build && pnpm run wait-until-postgres-is-ready && pnpm run wait-until-clickhouse-is-ready && pnpm run db:init && echo \"\\nDependencies started in the background as Docker containers. 'pnpm run stop-deps' to stop them\"",
+    "start-deps:no-delay": "pnpm pre && pnpm run deps-compose up --detach --build && pnpm run wait-until-postgres-is-ready && pnpm run wait-until-clickhouse-is-ready && pnpm run db:init && pnpm run reseed-bulldozer && echo \"\\nDependencies started in the background as Docker containers. 'pnpm run stop-deps' to stop them\"",
+    "reseed-bulldozer": "pnpm pre && bash scripts/reseed-bulldozer.sh",
     "start-deps": "POSTGRES_DELAY_MS=${POSTGRES_DELAY_MS:-0} pnpm run start-deps:no-delay",
     "restart-deps": "pnpm pre && pnpm run stop-deps && pnpm run start-deps",
     "restart-deps:no-delay": "pnpm pre && pnpm run stop-deps && pnpm run start-deps:no-delay",
diff --git a/scripts/reseed-bulldozer.sh b/scripts/reseed-bulldozer.sh
new file mode 100755
index 000000000..544d8f88d
--- /dev/null
+++ b/scripts/reseed-bulldozer.sh
@@ -0,0 +1,43 @@
+#!/usr/bin/env bash
+# Re-seed the bulldozer-js payments store from Postgres as part of restart-deps.
+#
+# Why this exists: bulldozer-js keeps its state in an on-disk LMDB store that
+# lives OUTSIDE Docker, so stop-deps/start-deps (which only reset the Docker
+# containers) leave a stale bulldozer store behind. After a fresh Postgres seed
+# that store is inconsistent, which is why we previously had to manually wipe
+# apps/bulldozer-js/.data and re-run db:backfill-bulldozer-from-prisma every
+# time. This script does that automatically: wipe the store, boot a temporary
+# bulldozer-js, run the one-way Postgres->bulldozer backfill against it, then
+# shut it back down again.
+set -euo pipefail
+
+PORT_PREFIX="${NEXT_PUBLIC_HEXCLAVE_PORT_PREFIX:-81}"
+BULLDOZER_PORT="${PORT_PREFIX}46"
+REPO_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+BULLDOZER_DIR="$REPO_ROOT/apps/bulldozer-js"
+
+echo "Wiping bulldozer-js LMDB store at $BULLDOZER_DIR/.data ..."
+rm -rf "$BULLDOZER_DIR/.data"
+
+echo "Starting temporary bulldozer-js on port $BULLDOZER_PORT ..."
+# Run the server directly (single node process, so we can reliably kill it) via
+# tsx's node loader — the same entrypoint the package's `start` script uses.
+(
+  cd "$BULLDOZER_DIR" && exec node --import tsx --expose-gc src/index.ts
+) &
+BULLDOZER_PID=$!
+
+cleanup() {
+  echo "Shutting down temporary bulldozer-js (pid $BULLDOZER_PID) ..."
+  kill "$BULLDOZER_PID" 2>/dev/null || true
+  wait "$BULLDOZER_PID" 2>/dev/null || true
+}
+trap cleanup EXIT
+
+echo "Waiting for bulldozer-js to accept connections on port $BULLDOZER_PORT ..."
+pnpm exec wait-on -t 60000 "tcp:localhost:$BULLDOZER_PORT"
+
+echo "Running Postgres->bulldozer backfill ..."
+pnpm run db:backfill-bulldozer-from-prisma
+
+echo "Bulldozer re-seed complete."