Skip maxed out queued DagRun candidates

[fat-catTW]

Why:

When _start_queued_dagruns() processes queued DagRun candidates, one candidate can fill the remaining max_active_runs capacity for a Dag or backfill. The scheduler already knows later candidates for the same (dag_id, backfill_id) cannot start in the same pass, but it still loaded the serialized Dag and re-checked the same limit for each one.

Solution:

Remember (dag_id, backfill_id) keys that reach their active run limit during the scheduler pass. Later queued DagRun candidates for the same key are skipped immediately, avoiding repeated serialized Dag loads and duplicate limit checks while preserving the existing max_active_runs behavior.

Was generative AI tooling used to co-author this PR?

• [X] Yes (please specify the tool below)

Generated-by: [Codex] following the guidelines

• Read the Pull Request Guidelines for more information. Note: commit author/co-author name and email in commits become permanently public when merged.
• For fundamental code changes, an Airflow Improvement Proposal (AIP) is needed.
• When adding dependency, check compliance with the ASF 3rd Party License Policy.
• For significant user-facing changes create newsfragment: {pr_number}.significant.rst, in airflow-core/newsfragments. You can add this file in a follow-up commit after the PR is created so you know the PR number.

[steveahnahn]

Nice work! cached_get_dag is keyed on the DagRun object rather than the Dag, so every candidate was a genuine miss with 3 get_dag_for_run calls down to 1 on your test.

One note: the two if active_runs >= max_active_runs log branches are nearly dead after this. get_queued_dag_runs_to_set_running already drops runs for Dags sitting at their limit, so they only fired when the limit filled mid-loop, which your new marking short-circuits. On your scenario main logs twice and this PR logs nothing. Worth either moving that line to where you mark the key, or dropping the branches.

Also, the backfill path got the same marking and has no test.

[fat-catTW]

Thanks, good catch. :D
I moved the logging into the shared key-marking path so the scheduler logs when a Dag/backfill reaches the limit and the remaining candidates will be skipped in the current pass. I kept the guard branches since they still cover concurrent/stale-count cases, but they now use the same helper.

I also added a backfill regression test for the same pruning behavior.

[potiuk]

Hello @fat-catTW - thank you for your contributions to Apache Airflow!

The Airflow community has introduced a limit of 5 open pull requests at a time for contributors without write access to the repository. You currently have 17 open pull requests, so - as a one-time step of introducing the limit - we closed the ones where maintainers have not engaged yet:

• Clarify Task Instance terminal state reporting #69482 - Clarify Task Instance terminal state reporting
• Restrict asset state store access to registered tasks #69764 - Restrict asset state store access to registered tasks
• Clean up stale KubernetesPodOperator pods #70140 - Clean up stale KubernetesPodOperator pods
• Skip maxed out queued DagRun candidates #71112 - Skip maxed out queued DagRun candidates
• Optimize scheduler backfill completion updates #71216 - Optimize scheduler backfill completion updates
• Isolate scheduled DagRun creation failures with savepoints #71287 - Isolate scheduled DagRun creation failures with savepoints
• Flush scheduled DagRun creation state promptly #71331 - Flush scheduled DagRun creation state promptly
• Allow paused Dags to finish running DagRuns #71697 - Allow paused Dags to finish running DagRuns
• Respect SMTP connection sender for task email alerts #72561 - Respect SMTP connection sender for task email alerts
• Reduce XCom requests when resolving mapped task inputs #72871 - Reduce XCom requests when resolving mapped task inputs
• Handle Dag run clear errors in task SDK client #72917 - Handle Dag run clear errors in task SDK client
• Deduplicate API response result conversions #73037 - Deduplicate API response result conversions

These pull requests stay open because maintainers are already engaged in them - they count towards your limit:

• Store executor callback DagRun references in the database #71214 - Store executor callback DagRun references in the database
• Limit Glue retry task UUID scan #71546 - Limit Glue retry task UUID scan
• Fix Dag run filtering for null start dates #71654 - Fix Dag run filtering for null start dates
• Allow API clients to refresh Dag bundles #72715 - Allow API clients to refresh Dag bundles
• Add missing Amazon provider unit tests #73169 - Add missing Amazon provider unit tests

This is not a judgement of you or of your changes. We never told contributors before that opening many pull requests at once was a problem, so there is nothing to feel bad about - and nothing is lost: your branches, commits and the review history stay where they are.

What we ask you to do is to make your first prioritization decision: choose which of the pull requests above matter most to you, and reopen them (up to 5 open at a time, including the ones still open) with the "Reopen pull request" button or gh pr reopen <PR_NUMBER> --repo apache/airflow. Reopen the ones you are ready to follow through - keep them rebased, respond to review comments and fix failing checks.

While your pull requests are waiting for review, the most valuable thing you can do is help in other ways - reviewing other contributors' pull requests, helping with issues, and taking part in the discussions on the devlist and Slack.

Why we introduced the limit, what it means for you and how to reopen or restore a pull request is explained in https://github.com/apache/airflow/blob/main/contributing-docs/32_open_pull_request_limit.rst.

---

Drafted-by: Claude Code (Opus 5); reviewed by @potiuk before posting