Why is it possible to run a thousand coroutines, while running a thousand threads is usually impossible?

The reason is that coroutines are lightweight threads managed at the application level (for example, using a scheduler in Kotlin or other languages), not system threads of the operating system. They do not require allocating a separate stack and OS resources, so thousands can be launched without significant load.

In contrast, system threads require allocating memory for the stack, context switching, and other resources, which limits their number. Coroutines, on the other hand, operate within one or several threads, switching very quickly and efficiently, allowing for scaling asynchronous operations without significant costs.