The outages aside, a couple of annoying, super-recent experiences:
- Our release process involves creating draft releases before QA, and publishing them when ready. We published a new release today and team members from the US were able to see it as published while members from different countries still saw the release marked as a draft and not visible publicly. We only noticed it two hours after making our announcements. We had to unpublish and publish it again to fix it.
- Recently, a colleague noticed a number of unrelated changes while reviewing a diff on GitHub, which led to us reverting a merge, only to later notice that the commit itself had no issues and it was the GitHub UI.
Outages and major issues are fine, they are loud and addressed directly. These kinds of issues go unnoticed, and they reduce trust silently.
I no longer trust the GitHub interface and actions. We had to re-test our release assets manually to make sure GH didn't mess that up.
Could this be yet another result of trusting AI code too much without proper review and testing? While discussing internally, one of my colleagues mentioned that "the whole internet feels fragile lately".
I'm curious to hear if anyone else has been facing such weird issues with GitHub.
