Apache Airflow version

2.10.5

If "Other Airflow 2 version" selected, which one?

No response

What happened?

After remove mapping tasks, subsequent normal tasks are not executed by the scheduler.
The issue was already described in #33164 which is closed with the note "Can't Reproduce".
I also encountered a similar bug, and prepared a small example DAG that reproduces the bug on the latest stable Airflow 2.10.5

What you think should happen instead?

Dag is executed successfully after increasing or decreasing the number of mapped tasks

How to reproduce

import logging
from airflow.models.dag import DAG
from airflow.models.variable import Variable
from airflow.decorators import task

with DAG(dag_id="example_map_removed_bug",
        schedule=None, start_date=None,
        tags=["example"]) as dag:

    @task
    def producer():
        "The first run generates a list [0, 1, ...] of 5 elements, each subsequent run produces 1 less element."
        value = int(Variable.get("test_value", 6))
        value = value - 1 if value > 1 else 5
        Variable.set("test_value", value)
        return list(range(value))

    @task
    def work(arg):
        logging.info("work task. arg = %s", arg)

    @task
    def finish(data: list[int]):
        logging.info("data = %s", str(data))
        logging.info("sum = %d", sum(data))

    t1 = producer()
    t2 = work.expand(arg=t1)
    t3 = finish(t1)

    t2 >> t3

• Unpause DAG, Trigger DAG, wait for all steps to complete successfully
  producer --> work[5] --> finish
• Do: Clear -> Clear existing tasks
  DAG starts over, this time the last step (finish) does not run, DAG Run goes into failed state
  producer --> work[5] (4 success + 1 removed) --> **F A I L**

Scheduler logs show error:

{dagrun.py:875} ERROR - Task deadlock (no runnable tasks); marking run <DagRun example_map_removed_bug @ 2025-04-04 21:25:51.774187+00:00: manual__2025-04-04T21:25:51.774187+00:00, state:running, queued_at: 2025-04-04 21:31:53.554772+00:00. externally triggered: True> failed

Operating System

Ubuntu 22.04.5 LTS

Versions of Apache Airflow Providers

No response

Deployment

Virtualenv installation

Deployment details

• Python 3.11.0rc1, virtualenv, Sqlite, airflow standalone
• airflow.cfg

[core]
dags_folder = /home/....../dags
donot_pickle = True
load_examples = False
max_templated_field_length = 16384

[database]
sql_alchemy_conn = sqlite:////......./build/airflow/airflow.db
load_default_connections = False

[webserver]
web_server_port = 1080
workers = 2
secret_key = .......

[logging]
base_log_folder = build/airflow/logs

Anything else?

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct