Update scripts/ingest_epoch.py: speed + balance + correctness fixes

scripts/ingest_epoch.py

@@ -394,6 +394,22 @@ def run_etl(
     # Step 4: Neo4j dump
     neo4j_path = dest_dir / NEO4J_FILENAME.format(epoch=epoch)
     if neo4j_path.exists() and not skip_neo4j:
+        # Auto-detect: if target DB is empty, use direct load instead of merge
+        # This prevents the merge-on-empty-DB hang when the first epoch uses --merge-neo4j
+        if merge_neo4j and not dry_run:
+            try:
+                _driver = _neo4j_driver()
+                with _driver.session(database=NEO4J_TARGET_DB) as _session:
+                    _count = _session.run("MATCH (n) RETURN count(n) AS c").single()["c"]
+                _driver.close()
+                if _count == 0:
+                    print(f"  [AUTO] Target DB is empty ({_count} nodes). Switching to direct load instead of merge.")
+                    merge_neo4j = False
+                else:
+                    print(f"  [AUTO] Target DB has {_count} nodes. Proceeding with merge.")
+            except Exception as e:
+                print(f"  [WARN] Could not check target DB node count: {e}. Proceeding with merge flag as-is.")
+
         if merge_neo4j:
             merge_neo4j_epoch_dump(epoch, neo4j_path, dry_run=dry_run)
         else: