test(settings): tighten 'preserves field when omitted' to actually prove preservation

The original test seeded the DB with the schema default (1) and asserted
the response was true after a partial PATCH. That passes both for
'untouched' and 'reset to default' — doesn't distinguish them. Now the
test toggles the DB column to false BEFORE the PATCH, then asserts the
false value survives both in the response AND in the DB row directly.

packages/server/src/routes/settings.test.ts

@@ -145,14 +145,27 @@ describe('PATCH /api/settings/instance — federatedRegistrationOpen', () => {
   });
 
   it('leaves federatedRegistrationOpen unchanged when omitted from payload', async () => {
+    // First toggle the DB column to false. If the field's value comes from the
+    // schema default (1) instead of the actual DB row, this test would still
+    // pass for the wrong reason. Toggling to non-default then asserting the
+    // non-default survives a partial PATCH proves real preservation.
+    testDb.update(schema.instanceSettings)
+      .set({ federatedRegistrationOpen: 0 })
+      .where(eq(schema.instanceSettings.id, 1))
+      .run();
+
     const res = await app.inject({
       method: 'PATCH',
       url: '/api/settings/instance',
       payload: { instanceName: 'NewName' },
     });
     expect(res.statusCode).toBe(200);
-    // Schema default is 1 → response should still report true
-    expect(res.json().federatedRegistrationOpen).toBe(true);
+    // Field still false (the partial PATCH did not touch it)
+    expect(res.json().federatedRegistrationOpen).toBe(false);
     expect(res.json().instanceName).toBe('NewName');
+
+    // Verify against the DB directly to rule out a response-shape-only fix
+    const row = testDb.select().from(schema.instanceSettings).where(eq(schema.instanceSettings.id, 1)).get();
+    expect(row?.federatedRegistrationOpen).toBe(0);
   });
 });