Downgrading DB leaves instance unable to clear tasks

### Apache Airflow version 2.11.0 ### If "Other Airflow 2/3 version" selected, which one? _No response_ ### What happened? Recently I accidentally applied DB migrations 28 through 68 while attempting to update from Airflow 2.10.5 to 2.11.0. After cleaning up the mess by downgrading the DB I noticed that I'm no longer able to clear tasks in order to re-try them. Trying gives you nothing except a lengthy exception in the `webserver` logs, the relevant part of which is this: <code>sqlalchemy.exc.IntegrityError: (psycopg2.errors.NotNullViolation) null value in column "id" of relation "task_instance_history" violates not-null constraint DETAIL: Failing row contains (nosify.notify_successes, cert_env, manual__2025-11-02T23:00:04.866429+00:00, 0, 1, 2025-11-02 23:04:02.810671+00, 2025-11-02 23:04:05.093361+00, 2.28269, success, 0, airflow-worker-0.airflow-worker.airflow.svc.cluster.local, airflow, default_pool, 1, default, 1, _PythonDecoratedOperator, null, 2025-11-02 23:04:01.165684+00, 720119, 68415, null, \x80057d942e, 2025 -11-02 23:04:05.146802+00, null, a96e4277-0d0e-4b53-868d-48e72e2dd9bb, null, null, null, null, nosify.notify_successes, null, 721182). [SQL: INSERT INTO task_instance_history (task_id, dag_id, run_id, map_index, try_number, start_date, end_date, duration, state, max_tries, hostname, unixname, job_id, pool, pool_slots, queue, p riority_weight, operator, custom_operator_name, queued_dttm, queued_by_job_id, pid, executor, executor_config, updated_at, rendered_map_index, external_executor_id, trigger_id, trigger_timeout, next_meth od, next_kwargs, task_display_name) VALUES (%(task_id)s, %(dag_id)s, %(run_id)s, %(map_index)s, %(try_number)s, %(start_date)s, %(end_date)s, %(duration)s, %(state)s, %(max_tries)s, %(hostname)s, %(unixn ame)s, %(job_id)s, %(pool)s, %(pool_slots)s, %(queue)s, %(priority_weight)s, %(operator)s, %(custom_operator_name)s, %(queued_dttm)s, %(queued_by_job_id)s, %(pid)s, %(executor)s, %(executor_config)s, %(u pdated_at)s, %(render…