Why is "upstream_failed" viewed as an intermediate state?

[pykenny]

Currently I'm reworking on the task lifecycle graph to get it up to the latest version of Airflow:

• Missing states in task state lifecycle diagram #40185

From definition of the state enums, task instance states can be grouped into three categories:

Terminal States

• success
• failed
• skipped
• removed (sort of, because a task instance can be reset to "none" state when it becomes available again during DAG run)

Intermediate States

• scheduled
• queued
• restarting
• up_for_retry
• up_for_reschedule
• upstream_failed
• deferred

Other States

• running (occupied by executor/worker)
• none (not processed by scheduler yet)

Other states seems to be reasonable in their corresponding category, but why is "upstream_failed" placed inside of "Intermediate States" category? A task enters this state when any of the required upstream tasks failed, and it won't be updated by the scheduler unless the failed states are cleared for another run.