Personally, I found the name of "Next Frame Event" is rather confusing too. From my understanding, what it do is acutally "Send event in current frame, Kick off next state's OnEnter event in current frame, make next state's OnUpdate start to execute in Next Frame". So the most confusing part is, if you put chains of "Next Frame Event" in multiple states, they'll all actually be executed in "Current Frame" instead of "Next Frame".
Is that really how it works? I haven't checked its code, but if it does it's quite powerful. Supposing you don't have any "every frame" you'de be able to chain as many events as you want (and, consequentially, their actions) in the same update. Not intuitive like you say, but still very powerful.
We're also toying with a project view editor window that shows FSMs and their connections visually, potentially giving you the benefit of hierarchical FSMs without complicating the FSM editor... it might be how we actually implement hierarchical FSMs.
This would be okay, but it wouldn't compare to the power of having a special state which, once clicked, instantly opened another FSM. You'd be able to save and re-use the current load templates without a problem like Jean requests. The only actual addition to the interface would then be back/next buttons (I'm thinking Flash/Illustrator here, pretty easy) for when you're digging into multiple FSM levels. The big functional change would be that, once the sub-FSM is called, the current FSM is paused and execution is resumed once the called FSM does its thing - maybe through a new/default "RETURN" event?
As for the current use of multiple FSMs for such tasks, I think we shouldn't downplay the extra hurdle demanded for having more than one FSM in the same object, never mind having to make sure the right actions are executed and the right event is passed to resume execution of the starting FSM. I'd rather keep my brain cycles for when it's really needed, ie. communicating between non-dependent FSMs. It's the same situation as using explicit messages instead of global variables - it's do-able, but it really defeats the main premise of ease-of-use and user-friendliness.
Just my $0.02