Github is claiming that a usage spike in 2026 is the cause of availability issues in 2025, so their explanation is clearly incomplete at best. The usage spike may be why things have failed to get better despite them putting effort into improving things, but it isn't the root cause of problems.
But the outages have been getting worse and worse even before anything related to AI took off.
The issue is that they're not a scrappy startup anymore, they are defacto running the internets development infrastructure and are owned by a trillion dollar company.
So the bar they're measured by has changed and they haven't even tried to keep up, paying lip service to reliability when you are critical infrastructure is not going to go well.
There were reliability issues in 2010 for sure, but it feels worse now; the period before acquisition was the most stable (2014-2017).
Funny how windows updates are never postponed for lack of "scaling". I know, I know, completely different stuff here - but arent test vms and ci vms being updated constantly?
Im old enough to remember the hotmail migration to win2k (then 2k3) and the postmortem. I was also old enough to look at the rotor source code. Yah, that one, running managed code in freebsd.
Their own greed is causing their issues. They could be doing a million different things to reduce demand, but they don't want to dampen their current growth and have opted to continue scaling up at the cost of quality.
Coupled with this (unsubstantiated but thorough) discussion on the internals of Azure, if even a fraction of this below-linked post is true, Github's abnormally-filesystem-intensive workflows would have wildly unpredictable performance and reliability forced onto Azure.
Azure also regularly has incidents due to capacity issues in several regions, so that many Azure-managed services also go down. Some of those incidents have been open continuously for many months now.
I think it doesn't need to be a large X% increase, just needs to hit some critical infra threshold where various services start failing and cascade. Weakest link and everything.
It's been on a downward trend before agentic coding took over. I suspect it's a mix of Microsoft culture and Microsoft infrastructure. It's starting to feel about the same quality as other Microsoft services.
Short aside, I have to rehost dotnet CLI binaries because their hosting infrastructure is so unreliable that it was causing CI failures regularly.
It began pretty much immediately after the acquisition. There was an uptime chart making the rounds a while back, and less than a year in, the all green data points of pre-Microsoft Github turned to lots of red. I assume brain drain, as everyone vested or otherwise completed their contractual requirements and cashed out. And, Microsoft has never had a great reliability culture in their cloud services, so no in-house talent to effectively take over.
Incomplete pull request results in repositoriesSubscribe
Update - We are actively reindexing the remaining ElasticSearch indexes. Our priority is ensuring correctness and avoiding further impact. We are taking a measured approach to safely backfill data and will share additional updates as progress continues.
Apr 28, 2026 - 15:58 UTC
Update - After yesterday’s incident, we are investigating cases where /pulls and /repo/pulls pages are not showing all indexed pull requests. This is because our Elasticsearch cluster does not currently contain all indexed documents.
No pull request data has been lost. As pull requests are updated, they will be reindexed. We are also working on accelerating a full reindex so these pages return complete results again.
Apr 28, 2026 - 14:51 UTC
Investigating - We are investigating reports of degraded performance for Pull Requests
Apr 28, 2026 - 14:17 UTC
GitHub took a massive hit in credibility when it got bought by Microsoft. We are a burned generation, we have seen the worst of Microsoft. This created a massive crack in the foundation of trust for most people.
Then Copilot happened. Some people dug how the training is done, and one GitHub employee responded by mail that every public repository including GPL repositories are included (the relevant Tweets are deleted unfortunately). The created crack has deepened. Some of us (incl. me) left GitHub.
As Copilot entrenched, Microsoft's product development practices and philosophy took over, and vibe coding started to be used by hordes of developers, GitHub's code foundations started to crumble. Add the big migrations they're doing & regressions they are causing on the UI now, and we're here.
GitHub's first enshittification cycle is over. Now we're starting the second cycle. The bloated, slow, entrenched hegemon's decay from relevance phase.
It'll be a slow decay. It won't fall in a day, but they golden era is long gone.
Any more context on the copilot training note? More pointers would be very interesting, but we'd need to keep in mind how many different underlying models were (are?) branded as copilot. I thought at some points the "copilot" model in autocomplete contexts was a finetuned GPT from OAI.
Re: GPL, there are other open access datasets of git repos that make some distinctions between copyleft licenses but those are older resources now.
Please see below. This is from the OG, "first generation" Copilot, from 2022. If I can find any more from my dusty trove, I'll edit or reply to this very comment. I can't do more digging now, because I'm in a pinch.
> Re: GPL, there are other open access datasets of git repos that make some distinctions between copyleft licenses but those are older resources now.
Arguably "The Stack" contains only permissively licensed code, but there are two repositories of mine inside it. One is a very simple logging library, without any license (which implies "All Rights Reserved"), and another is a fork of LightDM which I worked on, which is GPL licensed.
So any "permissively licensed" dataset probably contains at least one copylefted or strong copyrighted codebase, making them highly suspicious.
== EDIT ==
Found some. Kagi's date-constrained search to the rescue.
#2 makes #1 a big problem. AI-generated code is fine if you have thorough engineering practices around it. Are they blindly merging in AI generated code without review? Maybe. Thats an issue of engineering practices, not of the use of generative AI in general.
Well, it started just after the Microsoft acquisition, when AI did not exist, and coincided with news of Microsoft fully ingesting the GitHub team and forcing architecture and priority changes, and has steadily continued since. So idk, it’s a mystery. Maybe it was caused by the thing that did not exist when it happened. Microsoft just posted on a PR blog that it’s the thing that did not exist, and they’re famously truthful, open, and altruistic.
If you've worked with Azure you don't need to be explained what the problem is. I'll believe that workloads are different now than they were some years ago but... they literally are the cause for it so no sympathy from me there.
Azure is not the best, but it mostly works. GitHub gets only 98% reliability for git operation component, reading and committing. This is the most basic component. The fact they are not on this 24/7 and it isn't fixed is the result of a culture (=what is prioritized, what quality is accepted).
1. Increasing amount of AI-generated code in their codebase, decreasing the quality of the service.
2. Bought by Microsoft, and their bad engineering culture has spread to GitHub.
Perhaps it's a bit of both.