You are correct about the taking off part. If any of my ideas were to take off then I would need to swap out one of these services for an actual back-end.
Does it replace learning all of that server stuff at some point, no, of course not, but it allows us to focus on fewer issues at any given point in the course.
To the lock-in point, I've not seen that in what our students have built. There's a bit of unique code but mostly they're using standard AJAX & JSON techniques.
My own perspective is that Backlift and tools like it have real value for not just training but also for prototyping applications, rapidly evolving MVPs, etc. If something clicks and you've got to scale, great - that's when you invest in the real back-end. But there are also a lot of simple, low volume applications and websites where a back-end as a service tool could well be all that someone needs.
Additionally not all Backend as a Service have that extreme lock-in. In our case (iKnode) you build your backend using pure C# code, which is your and you can take with you. Creating Web Services out of the code you put in our servers is plain and simple.
I believe the customer pain is that for many applications, it takes many iterations before the product takes off -- and for non-developers, who have to hire developers to implement each iteration, it could be cost-prohibitive to get to the take off stage. Easy application prototyping and MVP developing products allows agility in fine tuning ideas to make sure it meets a need, before spending the money to hire a developer and implementing the final, more optimized version.
It's also advantageous for developers to use these products as well -- not only is it faster to develop (no need to learn a ton of APIs), it's also easier mentally to change the product to suit a business need (after all, it's easier to toss out an afternoon's worth of work than 2 weeks', if needed)