Ok, never thought of that usecase for moving parts. I mainly use it to test the relations of the parts to each other. But we can provide two tools or allow the user to decide if he wants to acknowlage constraints or not.
About the rotation: I know the issue. However, there are two points to it: First it's not trival to only solve for translations or rotations as all constraints have both as parameter in their equation, it's not possible to seperate both into seperate constraints. I have ideaas how to do it, but it's quite some work. Second, to handle the non-metric parameter space of a assembly system I had to implement some funky scaling algorithm and therefore rotations will nearly always rotate not around the parts origin but around some other point. Therefore even if I would seperate both, the result might not be as expected (for the case a simple translation is not enough).
To solve this in a userfriendly way a lot more is needed, and even if the infrastructure is mainly there it will take a long time to implement it. But a separation might result at least in a bit more expected results. I think about it ( I always prefere to design for the most general case, sometimes I need to be pushed for the obvious things