I don't know why, but stuff like this (e.g. the Apollo moon lander assembly code) is really interesting to me. I think it's because I wish I had this kind of knowledge and skill. It seems daunting to wrap my head around it and become proficient.
With the moon lander code, they surely had a very good specification of what it was supposed to do.
I worked with the second case once, some IoT-ish sensors that would be buried in the ground in greenhouses, to monitor soil data. The business logic was 90% specified, and of course, the remaining 10% took 50% of the time. Before you ask why we didn't do it in C at the very least, we had a very solid codebase from the previous products. Sure, we had to port things from a Toshiba microcontroller to an STM8, different architectures, but since we were working with 8 or 16 bits inputs it was kinda trivial to test every single possible input to make sure things matched.
My only previous experience with assembly at a Computer History college course where we coded for the PDP-11, the 6502 and a stack machine. So yeah, not a lot. Winging it while admitting you don't know what you're doing can get you decently far in some circumstances.
And don't underestimate the skill of putting together the entire product: being able to create behavior that is relatably live-like for millions of people on such a limited Plattform may not be Apollo-Level genius but is exceptional nontheless.
It would likely be much easier to use a more powerful MCU, at the per unit cost being higher.