This discussion has no need of the 3D symmetries defined in crystallography and elsewhere.[*] This is what most people will call a symmetry. This is the basic learning at school, and only a very few users know about crystallography...
The most basic idea of symmetry is left-right reflection as seen in nature ( many animals are symmetric ) and when you fold a piece of paper with wet paint on it. Most kids do that before they can pick up a pencil. Rotational symmetry is a bit more subtle.
That is what the symtool does in the case of a line and mathematical consistency would lead to doing this with a point. What the software does in the case of a point is either a 2 fold rotational symmetry or an inversion about the centre. ( These two are indistinguishable in the 2D case but differ in 3D ).
It's interesting what you say about pre-selection and order. Since each selection event can be trapped by the software , I don't see any fundamental reason why the selection order cannot be known but that may explain why this was implemented in an inconsistent way. It may require pre-emptively storing a linked list of selected items in the anticipation that some op. will be done on them and that order may matter.
The other fix would be to change post-selection to be the same as preselect but that will have habitual 'old school' users howling, mess up the doc and break compatibility with the original method. Not good.
Where is the group selection created and maintained if not in the software? Is this done by some GUI lib and then dumped in a random order to the caller?