Hi All, Indeed, markdown link checks are inherently flaky because they rely on external resources to be cooperative. This hinders PR progress.
Here's a recent example [1]: ERROR: 1 dead links found! [✖] https://medium.com/@jitenderkmr/demystifying-snowflake-ids-a-unique-identifier-in-distributed-computing-72796a827c9d → Status: 403 That link works fine in a browser. So +1 to remove markdown link checks (we can always fix when people report dead links). [1] https://github.com/apache/polaris/actions/runs/19534001072/job/55941993028?pr=2802 Cheers, Dmitri. On Thu, Nov 20, 2025 at 6:12 AM Robert Stupp <[email protected]> wrote: > Hi all, > > As Adam recently mentioned [1], the "Check Markdown links" workflow is > known to regularly produce false failures. > > It seems that some external sites have somewhat aggressive rate limits > and/or bot protections in place that lead to these false failures. > > We cannot control nor work around those externally controlled things. > PRs are getting blocked by these false failures. Retrying the workflow > does not help. And if the mentioned assumptions are correct, retrying > actually makes the (rate limit) situation even worse, leading to > other/more false failures from "Check Markdown links". > > I propose to remove "Check Markdown links" from the required checks > [2]. The workflow would still run, but not block PRs. Reviewers can > still inspect and cross-check potential failures from that workflow. > > Thoughts? > > Robert > > [1] https://github.com/apache/polaris/issues/3097 > [2] https://github.com/apache/polaris/pull/3102 >
