Not sure how to describe it but I see it like more of a approach and limitations or tech problem.

1- can't study it,in details, live with 0 risk of damage

2- can't test injecting data without risk of damage

3- the live data that we can extract/record (while monitoring input) with current tech is limited. I'm thinking that we're still at the stage comparable with using a multimeter to test a 8 layer pcb, with no schematics, while it's powered on.

I, personally, think science will get there at some point, but we'll need some serious tech advancements to do that.