Yeah in terms of CPU load I'm thinking more about the server handling hundreds of thousands or millions of user requests per hour, especially as it starts to get loaded and users start to get frustrated with slow pageloads and hit refresh multiple times.

The big websites have gotten so good at handling large amounts of users that hardly anybody remembers what using video services on the internet was like in 2010, even if they lived through it, and nobody has any patience anymore.

Imagine this scenario:

A Peertube admin rents a VPS from a reputable host provider, and runs a Peertube instance on it. Some months later, a video on that instance gets some attention and a link to it gets passed around some larger social media platforms. Two hundred thousand people try to watch the video in less than an hour, hitting the server with requests from all over the planet.

Because the admin has chosen to host on quality hardware in a well-run data center rather than on a used laptop in a shoe closet, the server doesn't fall over. The provider automatically scales the CPU and bandwidth to meet the increased demand, and scales the pricing tier accordingly, and the admin forgot to set any usage caps when they set up the VPS because they never expected this kind of workload. The instance still chugs because it's just a single VPS, not an entire server farm, it was never built for this, so after the first ~50,000 requests page load starts to fail, and more and more users get frustrated with buffering issues because they're used to YouTube, and they reload the page multiple times creating an exponential number of requests and then give up and leave.

Because of the operational problems very few of the new users become regulars, and only 5 of them donate a little money to the admin, who gets stuck with a $500,000 invoice from the host provider for the usage spike.