When I took Intro to Computer Science and Programming on edX from MIT (The original 16 week 6.00x before they broke it up into two courses), they broke up the short videos with "finger exercises" which was like the interrupting questions on Udacity, but there were more of them and they were a lot more comprehensive. It was worth enough of your grade so there was motivation to do them, but not so much that you couldn't skip them if you felt you already really knew it. That was, to date, the best MOOC I've ever taken.

I agree that Coursera can sometimes feel a bit too much like copy/pasting a college class onto the internet, but it really does vary a lot by course. For example, Robert Ghrist's Single-Variable Calculus on Coursera was amazing. 15 minute animated video lecture followed by 10 problem homework assignment.

As far as the scheduled vs self-paced difference, there are ups and downs to both. I have fallen behind in a class before and then abandoned it because I missed a deadline. But knowing that "now is your chance" to take a course can be more motivating than doing self-paced sometimes. Deadlines can be useful.

I really don't know what's best, but I'm a huge fan of the open education movement and I see innovations happening all the time. For example, each course in Coursera's Data Science track has a "due date" for full credit, then a "hard due date". Each day between them, your score on that assignment loses 10%. You have a total of 5 late days to apply throughout the course. That's enough to save you if you fall off the wagon for a bit and knowing that you're losing a bit each day can motivate you to get it done, while being unable to submit after missing the first "due date" can make you want to quit.