I get your meaning, but additional abstraction layers are equally dangerous because they obsfucate underlying constraints and result in incorrect reasoning abou

I get your meaning, but additional abstraction layers are equally dangerous because they obsfucate underlying constraints and result in incorrect reasoning about a system. There's a reason RTOSs don't have very tall software stacks with multiple layers and virtual machines between them and the hardware. When your garbage collector invoking at the wrong millisecond 0.01% of the time means everyone on the craft is incinerated in the atmosphere... that's a different level of reliability requirement.