The problem, though, is something along these lines: If you're doing an action in a cutscene, it's going to be something that you don't normally do. Resident Evil 4, for example, does not have any button that you can press to kick a zombie in the head, except during the kicking-a-zombie-in-the-head cutscene. How is the player supposed to know what button kicks the zombie in the head, when it is such an infrequent occurrence that it needs a cutscene dedicated to the event? It can work in situations where the flourish is based on something a known button already does - pressing the sword button to perform a fancy sword move, for example - but how do you present players with a variety of unprecedented options without telling how to perform them?
Obviously, the huge controller glyphs are ugly. However, this is still fundamentally a problem of user input. That means it's not specific to games, but is pervasive throughout software design. Let's take a page from the book of standardized Windows form menus. These are a handy example of an input cue that is both subtle and explicit. Notice how one letter is underlined? Press Alt to focus on the menu (a common enough standard that it is usually not shown at all), then press that underlined letter's key. There's no big ugly "Alt-F" tooltip hovering over the File menu, though. No obtrusive, ugly, patronizing reminder. It strikes an excellent balance between subtlety and clarity.
Make the cue too vague and your player will have no idea what to press, if they don't do it normally. Make it too bright and you get articles like this one. What we need to look for are elements that do it right. I think color-coding is an avenue that could be explored, personally - except that color seems to be vanishing from consoles entirely. Handhelds have it easy, since you can refer to the positions of the buttons relative to the screen, but there's a woeful lack of standards to determine what the face buttons do, once you have more than two of them. On the Wii, you could highlight an area of the screen to move the cursor to and perhaps press a button, but that only deals with pointing, not waving....
This could just be boiled down to a game design issue. If you're making your player do something that they have not done recently, then you have to tell them how to do it. If giving them explicit instructions is bad and subtler instructions are infeasible, then the only remaining option is to not make the player do anything that they haven't already done. I should clarify - something that they haven't already done in that game. I don't know if that's the best answer... but software design is about compromises, and in this case we can treat games as software.