These kind of booleans is what Ogre does / used to do and they’re quite nasty, as they’d fall directly on the fast path required to calculate global matrices.
If possible, I suggest keeping a non-scaled parent hierarchy, and adding a separate child for the scaled parent model.