ShortCircuitOperator not short-circuiting as expected with expanded tasks

[Chais]

Here's my DAG:

import random
from typing import List

import pendulum
from airflow.decorators import dag, task


@dag(
    "test_foo",
    schedule=None,
    start_date=pendulum.now(),
    render_template_as_native_obj=True,
    dag_display_name="Test random things",
)
def test_foo():
    @task.python
    def start() -> List[int]:
        return [i for i in range(10)]

    @task.short_circuit
    def step_one(i: int) -> bool:
        print(f"Hello from task {i}")
        return random.random() >= 0.5

    @task.python
    def step_two(do: bool):
        print(f"{"D" if do else "Not d"}oing stuff")

    step_two.expand(do=step_one.expand(i=start()))


test_foo()

This results in 21 tasks total, as expected. My issue, or misunderstanding maybe, is that in step_one if the output is False the ShortCircuitOperator logs No downstream tasks; nothing to do, even though there clearly is one for each task.
How do I get this to behave as expected?