I voted for stacked window. It's about interaction reduction, displaying the show-notes is not intrusive, and it reduces the amount of actions required.

As it is, if you want to view show-notes and then download, it's 5 input actions. Select episode, view show-notes, return, select episode, play.

With the proposed method it's reduced to 2 possible input actions. Select episode, show-notes are displayed, select play.

It's more efficient and streamlined, and more importantly - it's not invasive at all, as displaying the show-notes is completely passive and does not require any input or thought from the user.