The difference between suspend fun doSomething() and fun CoroutineScope.doSomething() is one of the key points of structured concurrency.

A suspend fun doSomething() guarantees that it will always complete all work before returning. By using the coroutineScope builder it can launch “internal” coroutines that are not exposed to the caller. A caller can be sure that no new work is running at the time doSomething returns.

A fun CoroutineScope.doSomething() is passed a scope in which it may launch a new long-lived coroutine. A caller should expect that calling this version of doSomething will create a new coroutine in the passed scope that lives longer than the call of doSomething. This is mostly not needed in production, but is available when you want to build custom coroutine builders (e.g. alternatives to async).