I've thought about this issue quite a bit because I had a use case (using scala) with lots of graphical data structures and GC mark phases were totally killing my performance, and I thought about using rust until I found out my graph shenanigans wouldn't work there either.
One of those hairbrained research-y ideas I have bouncing around in the back of my head is to fix this. The thing about graphical structures is that there is a huge body of work in graph theory that could be used to provably and deterministically cover >90% (ballpark) of these data structure use cases, but likely at the cost of compiler performance. And if I were even remotely competent with rust macros, I would totally hook up some annotation macro that would call out to Z3/CVC4 to determine satisfiability and then rewrite the code safely (even if it uses unsafe blocks under the covers) or throw an error. But thats for another time :)
One of those hairbrained research-y ideas I have bouncing around in the back of my head is to fix this. The thing about graphical structures is that there is a huge body of work in graph theory that could be used to provably and deterministically cover >90% (ballpark) of these data structure use cases, but likely at the cost of compiler performance. And if I were even remotely competent with rust macros, I would totally hook up some annotation macro that would call out to Z3/CVC4 to determine satisfiability and then rewrite the code safely (even if it uses unsafe blocks under the covers) or throw an error. But thats for another time :)