I think it’s a reasonable solution still, although you’ll probably want to disable freezing globally as the native side will still try to freeze data. I can imagine an implementation designed for the new memory model would be better, but the current stately one should be serviceable assuming the performance constraints aren’t extreme