It's interesting to see Twitter down, apparently worldwide. It's obviously sections of functionality that are having difficulties; unfortunately the core ones.
I remember all the videos about "how Twitter engineered for scale", as far as I can see, Twitter uses a mix of it's own data centres, AWS and GCP. There are multiple blog posts about wonderfully engineered solutions, with Service Meshes, Microservices, Kubernetes, etc. etc.
There is rate limiting taking place (although it's probably not wise to have the default message "Something went wrong. Try Reloading."), some subsystems are working, so it seems they have tried very hard to build in some best-practice for availability.
But every system will have problems, there are limits, even when you have the Richest Man In History as your CEO...
It's not yet clear the underlying cause.
The issue I would fault Twitter on, is lack of communication / feedback - after three hours, it's status page shows all systems working, and there has been no messaging from the company that the system is encountering difficulties. This is snowballing, as more people rapidly click "refresh", then taking to other social media to ask the same question thousands of times - "is it just me?"
I'd certainly add communication of outages to your checklist for availability.
#microservices #availability #communication