fix(models): cascade one-to-one child rows on application purge

Purging an app from the Recycle Bin crashed with IntegrityError
(NOT NULL constraint failed: container_scale_policies.application_id)
when the app had an auto-scale policy: with no delete cascade,
SQLAlchemy nulls the child FK on parent delete, and these one-to-one
rows declare it NOT NULL. Scale policy, sleep policy and wp_site all
share the shape; all three now ride the purge.

Reported on Discord (vaultwarden/hedgedoc delete-reinstall flow).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

backend/app/models/container_scale_policy.py

@@ -23,7 +23,10 @@ class ContainerScalePolicy(db.Model):
     last_scaled_at = db.Column(db.DateTime)
     updated_at = db.Column(db.DateTime, default=datetime.utcnow, onupdate=datetime.utcnow)
 
-    application = db.relationship('Application', backref=db.backref('scale_policy', uselist=False))
+    # cascade: application_id is NOT NULL, so a real delete of the app (Recycle
+    # Bin purge) must take this row with it — the default null-out IntegrityErrors.
+    application = db.relationship('Application', backref=db.backref(
+        'scale_policy', uselist=False, cascade='all, delete-orphan'))
 
     def to_dict(self):
         return {

backend/app/models/container_sleep_policy.py

@@ -17,7 +17,10 @@ class ContainerSleepPolicy(db.Model):
     asleep = db.Column(db.Boolean, default=False, nullable=False)
     updated_at = db.Column(db.DateTime, default=datetime.utcnow, onupdate=datetime.utcnow)
 
-    application = db.relationship('Application', backref=db.backref('sleep_policy', uselist=False))
+    # cascade: application_id is NOT NULL, so a real delete of the app (Recycle
+    # Bin purge) must take this row with it — the default null-out IntegrityErrors.
+    application = db.relationship('Application', backref=db.backref(
+        'sleep_policy', uselist=False, cascade='all, delete-orphan'))
 
     def to_dict(self):
         return {

backend/app/models/wordpress_site.py

@@ -100,7 +100,10 @@ class WordPressSite(JsonColumnMixin, TimestampMixin, db.Model):
     # Timestamps
 
     # Relationships
-    application = db.relationship('Application', backref=db.backref('wp_site', uselist=False))
+    # cascade: application_id is NOT NULL, so a real delete of the app (Recycle
+    # Bin purge) must take this row with it — the default null-out IntegrityErrors.
+    application = db.relationship('Application', backref=db.backref(
+        'wp_site', uselist=False, cascade='all, delete-orphan'))
     environments = db.relationship(
         'WordPressSite',
         backref=db.backref('production_site', remote_side=[id]),

backend/tests/test_application_soft_delete.py

@@ -157,6 +157,36 @@ def test_purge_is_what_removes_the_volumes(app, dead_app, monkeypatch):
     assert calls and calls[0].get('volumes') is True
 
 
+def test_purge_takes_the_one_to_one_policy_rows_with_it(app, dead_app, monkeypatch):
+    """Field report: purging an app that had an auto-scale policy died with
+    `IntegrityError: NOT NULL constraint failed: container_scale_policies
+    .application_id`. Without a delete cascade SQLAlchemy's default on parent
+    delete is to NULL the child FK — and these one-to-one rows declare the FK
+    NOT NULL, so the purge itself crashed. The policy rows must ride along."""
+    from app.services import docker_service
+    monkeypatch.setattr(docker_service.DockerService, 'compose_down',
+                        classmethod(lambda cls, path, **kw: {'success': True}))
+    monkeypatch.setattr(application_restore.CronService, 'clear_application',
+                        classmethod(lambda cls, _id: 0))
+
+    from app.models.container_scale_policy import ContainerScalePolicy
+    from app.models.container_sleep_policy import ContainerSleepPolicy
+
+    with app.app_context():
+        db.session.add(ContainerScalePolicy(application_id=dead_app.id))
+        db.session.add(ContainerSleepPolicy(application_id=dead_app.id))
+        db.session.commit()
+
+        ok, warning = recycle_bin_service.purge('application', dead_app.id)
+        assert ok
+        assert warning is None, warning
+        assert Application.query.get(dead_app.id) is None
+        assert ContainerScalePolicy.query.filter_by(
+            application_id=dead_app.id).first() is None
+        assert ContainerSleepPolicy.query.filter_by(
+            application_id=dead_app.id).first() is None
+
+
 def test_purge_honours_an_explicit_remove_data_false(app, dead_app, monkeypatch):
     """Database engines default to preserving their volumes — losing the data is
     the only irreversible part of that uninstall."""