My hunch is that’s been proven to be a Turing completeness level challenge (ie I think you’d have a tough time even coming up with a way to compute a probabilit

My hunch is that’s been proven to be a Turing completeness level challenge (ie I think you’d have a tough time even coming up with a way to compute a probability that it’s non-tampered). I suspect even that heuristically will be exceedingly hard to do for an optimizing compiler since the optimizations it performs are very complex data crunching to detect various invariants that allow for a safe optimization (eg notice how when you debug optimized native code it jumps around between random lines?) A stochastic approach might give you something but you’d need to actually explore the idea to understand what that is if anything. I’m not aware of anything in that space but I’m also only a casual observer and not that read up on the literature.