Hi, do you have a memory insights trace that you could share? Or even just a full call stack to see where that call to CopyHierarchy is coming from?
Copying the rig hierarchy can be an expensive operation - the more complex the hierarchy, the more memory is going to be required and the longer the operation is going to take. So the general way to optimize that is to reduce the complexity of the rig hierarchy. Or possibly run at a reduced mesh LOD that has fewer bones & curves, and so could use a control rig with a reduced hierarchy. But that’ll impact on fidelity which is why I’d be curious to see exactly where that call is coming from to see if there is anything else that can be done.