James Addison

This test flakiness has [reappeared](https://github.com/sphinx-doc/sphinx/actions/runs/8251553945/job/22568877990?pr=12065#step:6:2150); reopening this issue.

What would be useful here is a way to reliably make the test failure occur. So far I've found it difficult to determine whether an attempted fix like the one...

I wonder whether https://github.com/python/cpython/issues/85307 could be related to this.

I think there's a chance that #12126 may resolve this, despite not being expressly intended to.

> I think there's a chance that #12126 may resolve this, despite not being expressly intended to. We have not yet merged that pull request (twenty-six). Meanwhile, I have not...

This test flakiness has not re-occurred within the past two weeks, and I believe that the test is now reliable. _In theory_ I think that we could now safely revert...

> I'm not fan of the configuration name but I don't have a better one Yep, it took me a while to decide, and doesn't seem perfect to me either....

I think that merging this would hide the problem reported in #12038 -- so I think we should try to figure out the problem there beforehand.

> I think that merging this would hide the problem reported in #12038 -- so I think we should try to figure out the problem there beforehand. ...with one exception...

No, sorry, I'm talking nonsense. The affected test doesn't _only_ check the link status, it _also_ checks for a specific SSL-related failure message. False alarm.