vffgaston:
I'd, instead, pose a single idea: the hardware growing power (it doubles every two years?)
And this is exactly what has made Python in embedded systems possible. There is no way Python would ever be practical on an ATmega328P but on an STM32, ESP8266, ATSAMD21, etc. board you have plenty of memory and clock cycles to fit a reasonable sized project and an interpreter even if the code is not as efficient as it could be.
For a hobbyist, spending an extra $1 for a microcontroller with the extra capacity to allow for an easier programming experience is no big deal. For a mass manufactured product it will be worth paying the salaries of the programmers for the extra time to write efficient code that can fit on a slightly cheaper microcontroller.
vffgaston:
how many projects are 'time not critical' ones (compared to the ones in that a uS makes the difference?).
Power efficiency is becoming increasingly more important as so many embedded devices are trying to do a lot on battery power in a small package. Maximum power efficiency comes from efficient code (every clock cycle while awake counts). Code written in Python will never be as efficient. This is an important factor even for some hobbyist projects.
So to me it seems like Python is good for a hobbyist but will rarely be the choice for a professional embedded systems programmer. Is it really worth it for a professional to learn an additional language just for the rare cases where Python does make sense to use? The thing I really like about Arduino is that, even though it's meant for beginners and hobbyists, the knowledge gained could be applied directly to a professional career if that's where it takes you. If you get started with Python and then decide to embark on a career in embedded systems programming all your Python programming skills you learned will not be used and you're going to need to learn C/C++ instead.