Loving Bitlash, in that it provides a means of transitioning from interpreted shell to hosted code. Always a little worried about being caught in a cul-de-sac though, even if it's a garden of eden 
That's one of the reasons for raising this query to begin with.
dxw00d understand what you're saying on Firmata support. If in practice the community is turning away from it we have to take this into account. We're not using Megas (or Arduinos for that matter), so I'm not too worried on that front. However, I don't believe it's easy to substitute and write your own protocols, at least for the average person.
The I.T.A. parallel from Mike was pivotal to me understanding where I'm coming from. It clarified in my mind what exactly I feel needs to be taught. I think you're right if the aim is to get them to use C, then it's a bit mad to teach them Python. However, the aim is to teach them to be programmers, not C programmers (and importantly NOT computer scientists, this is like the difference between a carpenter and a dendrologist).
I'm hoping to teach the fundamental skills which transcend any language and are a matter of the programmers craft, insights like the concept of a variable, the habits and shortcuts of how and when you structure data into sequences (containers for items without names = list/array) or dictionaries (containers for items with names = struct/object) depending on what you're trying to achieve, sensible approaches for choosing names, how to break complex behaviours down into smaller parts. I believe these aspects are easier to learn in an interpreted, untyped shell language, and especially one which can be used for so many jobs, but in any case, these skills can transfer. In the ideal world, session participants will have developed some of the skills of a pragmatic programmer, concepts which can be used across javascript, java, ruby, php and so on.
If the fishing trip is successful, they bite the worm, and become developers probably C is one of the least useful languages for them when branching out into web programming, audio, graphics. If they were stuck with python, it could be worse - it can be used for almost anything and this is where the parallel with I.T.A. breaks down, which really was a cul-de-sac, and could never be used for real-world tasks.
I have to differ with dxw00d that you can explore individual concepts in C. I thought it was a matter of fact that causing anything to happen through Arduino C demands a complete program compiled to hex, which means you're exposed to void, function defs, function calls, (implicitly) library calls and many other things before you can achieve anything at all. I'm talking from a semi-religious point of view here, as I've been convinced by Constructivism... Constructivism (philosophy of education) - Wikipedia and indeed Constructionism... Constructionism (learning theory) - Wikipedia , which takes things a little further, and more literally. Both of these emphasise being able to act on your world to figure stuff out or else most people just won't figure it out. If you can go pin.write(1) and see the light come on, this meets the challenge. If you have to write the whole of Blink without errors and upload and run it to confirm your comprehension of 10 concepts at once, this is a real problem.
It's no surprise that Papert (the original Constructionist guy) was behind the family of programming tools leading to Scratch. This could go some way to explain your experience of educationalists too, Mike. This atomic, active manipulation is the sort of experience they feel individuals need to have in order to make sense of programming. The geeks amongst us can cope even in the face of complexity. I fear many people other people are left by the wayside because our typical teaching approaches mean they cannot build up incrementally. I'm probably reaching more for the non-geeks here, but as a geek myself I'm badly qualified 
Has anyone in the thread had success teaching non-programmers with either C or python or anything else? Be good to hear your anecdotes of how key concepts 'dawned on' people. I'm especially wondering if anyone's had success converting people who are not classically interested in computers/logic into using programming for its practical benefits? That's the particular line I'm trying to go down. Happy to teach natural geeks too, but I'm wondering if I can achieve some evangelical programming; on a mission