You have to design for scalability. Bottlenecks may be wherever. Even if their virtual server CPU and RAM can scale up, other stuff may be bottlenecks. Maybe the connection to the DB. Maybe the DB is elsewhere and doesn't scale. Can't really reasonably guess from the outside.

Mass cancellation is not usually a thing they would design around bottle-necks. It also doesn't add value to them.

Scaling has a budget, I’m sure. They’ll only pay for so much.

If your page is just static, e.g. no login, no interaction, everyone always sees the same thing then it scales easily. Scaling means you copy the site to more servers. Now imagine a user adds a comment. Now you need to add the comment to every copy of your site, so that everyone sees it regardless of which server they use. So a comment creates more work the more servers you use. And this is where scaling becomes a complex science, that you need to manually prepare for as a software developer. You need to figure out what data will be stored where and accessed how.

It could also be poor graceful failure. What we see as a crash may be from some unavailability deep in a long pipeline of services.

I feel like Disney has internal stuff? I listened to a podcast where an ex employee changed the fonts on a bunch of stuff to be wingdings, etc, and made everything unusable.

Side tangent, seems odd to me this is still a thing. Most company websites aren't hosted on premises, so do these services like (i assume) AWS not scale for when there's traffic?

Scaling is only for companies that have not been allowed to purchase and enshittify every serious competitor. (Pixar, Marvel, HBO...)