Julia was designed specifically to have language semantics that would make compiler optimizations as easy as possible without sacrificing usability.
Bolting an efficient, all-language compiler onto Python post-hoc is essentially impossible without fundamental changes to the language.
It is getting tired by now.
I'm well aware that you can optimize code in a highly dynamic language (I'm mainly a julia programmer afterall).
The problem with Python isn't dynamism, it's the way the dynamism was baked into the language, and the very leaky way it affects the language semantics.
As mentioned, imagine having something like the Smalltalk becomes:, where in a flip second, all occurrences of type A turn into B, across the whole application, including already JITed.
Or stopping the application at any given moment, recode something, hit continue, and the application carries on, as if it was the original implementation to start with.
Or given they were full workstation OSes, apply such dynamism to the kernel level code.
Do you think that people just haven't tried hard enough?
Facebook was interested into doing the work for PHP.
Shopify is interested into doing the work for Ruby.
Google was interested for a moment, but then instead of spending money on Unladen Swallow, they decided to port Python code into Go instead.
NVidia is interested, but only to the point it is a JIT for CUDA workflows.
Now finally Microsoft and Facebook, seem to be willing to invest into Python, so lets see if this time how long they will care.
Interest alone doesn't make it happen, money, people and management willing to wait, make it happen.
GraalVM would never have happened, if Sun or Oracle had killed MaximeVM after almost a decade of research work.
Shopify's JIT is showing good enough performance, that they are rewriting C code into Ruby.
This is the thing of Python culture versus everyone else, rewriting in C is considered "Python".
As far as I understand, in practice, JavaScript code relies less on foreign function interfaces to code that's written in other languages (because you can't reliably do that in a browser). This point makes it so that important JavaScript packages tend to be written in JavaScript, whereas important Python packages tend to be written in C or Fortran or Rust or whatever. If a JIT compiler for Javascript changes how FFI works, then that's less big of a deal.
Again though, not very knowledgable about JavaScript, so if someone wants to come and correct me, or offer additional context, I'd be very appreciative.