Busy Beaver optimization?

[Rudxain]

Note

I selected "General" rather than "Question" because this isn't entirely a question about HVM per-se, and because I have no idea what the proper category would be

I was reading this computer-science/mathematical paper, and thought "It's very likely that their Laconic-to-Turing-Machine compiler doesn't optimize code pretty well, thereby explaining the immense upper-bound in the number of states".

So I had an idea: What if the Laconic source-code were rewritten in HVM-Lang, then optimized by the HVM compiler, then finally re-compiled to "Turing-Machine byte-code"? That way, we would very likely get an upper-bound that's closer to its global minimum (if optimizing for space, not time).

My question is, is this even possible? Would the transpilation cancel all the optimization benefits?

AFAIK, HVM isn't "magic". A TM is an inherently stateful mutable system. But perhaps we could exploit the strong locality-of-reference caused by sequential (in the HDD way, not ROM way) memory-access.

BTW, I'm sorry if this is the wrong place to ask these sort of questions