GitHub has reached the part of platform adulthood where it starts making speeches about resilience right after it ruins your day.
That sounds harsh, but look at the week it just had: merge queue regressions, search-backed pages coughing up empty results, Actions delays, and the kind of incident log that reads like somebody shook a server rack and heard apologies fall out. GitHub’s own post said the April 27 search mess was probably a botnet overload. Their own CTO said they now need to design for 30x today’s scale because agentic workflows have gone feral since December.
Thirty times.
That is not “a little growth.” That is a platform getting mugged by the future.
Here’s the part that should make everyone deeply, personally annoyed: Git itself was fine. The raw API was fine. The CLI was fine. The smug little browser layer—the place where most humans actually work—was the thing melting into soup. Which means the most dependable path through GitHub is the one that looks least like the product everyone worships. gh pr list kept standing there like the only adult in the room while the website had a panic attack.
And honestly? That feels about right.
The modern developer stack loves pretending the nice web UI is the center of gravity. It’s not. It’s just the decorative topsoil over a swamp of indexes, queues, auth, caches, background jobs, and whatever other cursed machinery keeps the whole circus from collapsing. Once one of those layers sneezes, your “collaboration platform” turns into a pretty error page with branding.
GitHub’s official availability post is basically an accidental confession. They say availability first, then capacity, then new features. Good. Wonderful. Love that for them. But if the platform now has to re-architect itself around the blast radius of agentic nonsense, bot pressure, giant monorepos, and a firehose of PR activity, then the message is pretty simple: the product has outgrown the manners of its own infrastructure.
This is what happens when a tool becomes the default place where everybody does everything.
Code review. Issues. CI. Artifact delivery. Notifications. Security scanning. Merge queues. AI agents. Project boards. Search. Comments. The whole mess is one giant dependency stack wrapped in a blue coat and a cheerful brand voice. When it works, people call it productivity. When it doesn’t, it’s a very expensive reminder that centralized convenience is just fragility with better typography.
I’m not even saying “leave GitHub tomorrow” like some solar-powered hermit with a Forgejo tattoo. I’m saying stop acting surprised when the all-in-one super-platform behaves like an all-in-one super-platform: overloaded, leaky, and one bad rollout away from turning your afternoon into an outage-themed improv show.
If your team’s shipping process dies because search is stale or PR pages are missing results, that’s not a Git problem. That’s a “you built a cathedral on a Jenga tower” problem.
The funniest part is that GitHub knows this. They wrote the post. They named the failure modes. They admitted the isolation work wasn’t done yet. They admitted some systems were still single points of failure. They even told you which parts survived the blast, which is a cute way of saying, yes, the plumbing worked while the chandelier fell off the ceiling.
So no, I do not find the “we hear your pain” part comforting. I find it bureaucratically charming. The pain is the product now. Everyone gets a little. Maintainers get it, contributors get it, and the poor bastard reviewing a PR at 2 a.m. gets it when the page loads with the confidence of a dead printer.
Build your escape hatches.
Mirror the repo. Know the CLI. Know the API. Keep your CI from worshipping a single vendor like a cursed household shrine. Because the moment your workflow depends on GitHub behaving perfectly, every outage turns into a personal insult. And GitHub, bless its corporate heart, seems determined to make that a recurring hobby.
Maybe one day it really will be boring again. Today, it’s weather.
And weather doesn’t care about your sprint.