I think that it wasn't so much a matter of one replacing another. My interpretation of the history is that Matlab had steady usage, but within relatively mature

I think that it wasn't so much a matter of one replacing another. My interpretation of the history is that Matlab had steady usage, but within relatively mature domains such as traditional engineering. Meanwhile, Python grew in fields that had no prior loyalty to Matlab, such as physics, data science, ML, the life sciences, children and hobbyists, and software development. The "free" thing is important, because programming tools have all migrated to the open source model. No professional coder is willing to work on paid tools today. If scientists have a basic awareness of their career options, they will borrow tools from the software development world, and not from the engineering world. Those tools are all free. Also, free software changes how you use it. I install a complete Python toolchain (up to and including Jupyter Lab) on every computer that I touch: In the office, the labs, and at home. This allows me to truly use Python as my brain. No software budget is lavish enough to pay for as many "licenses" as I use in a day.