AWS usually advertises performance and scalability for their services. For example, this says that their Network Load Balancer "is designed to handle tens of millions of requests per second while maintaining high throughput at ultra low latency". All they seem to quantify for the Global Accelerator is latency.
In terms of requests per second or open connections, how well does a Global Accelerator scale if used for http traffic?
While a single region means you don't fully use GA features. Whether it gives you a (practical) speed boost depends on where the (only) deployment region is relative to the majority of your end users.
You can get a feel for this using https://speedtest.globalaccelerator.aws
For example, if my deployment is in Sydney and my end users are also in Syndey(I'm running this from Sydney), you can see in this screenshot it's a "meh":
This also shows if for some reason my deployment is in the US while my end users are in Sydney, it actually helps. Same with deployment in Sing and users in Sydney. Though, normally if it's a single region thing it would naturally be deployed where the users are based, I think.