Due to a mistake in my testing setup, I included a mistake in my article on async functions.
The mistake is I wrote that if we have a non-async function that spawns a Task, we depend on the callsite's actor isolation to determine whether the new Task runs on the main actor.
Turns out that's not the case. I observed that the spawned Task is never main actor isolated even when callsite is.
I assume that that's because there's no actor isolation applied to the function.