Improved Output Node Behavior in the Node Editor

I've noticed you can duplicate the group input node and distribute it on the tree at various sockets you want to drive as parameters from a higher level (or directly from the modifier panel). I don't know if you're supposed to be able to do this or not, but I know I love it because it helps keep the noodle situation under control in a really nice way!

The docs say additional input and output nodes are interpreted as "duplicates". Since duplicate input nodes behave as described above, it seems reasonable that duplicate output nodes would too.

However, this currently does not seem to be the case. When I try the same thing with the output node, It's as if it only recognizes one output node at a time. In fact, once a duplicate output node has been introduced to the tree, getting it to switch between which output node is being recognized becomes very difficult, and sometimes results in a crash.

The docs also say that the node tree is evaluated backwards. Even if that's so, I don't understand why I cannot use multiple output nodes in this way. It seems duplicate output nodes could simply be altered so that they don't get evaluated, and instead only contribute to the original output nodes' argument population.

The ability to run parameters directly to the output node the way I've described would make for exquisite readability. As someone who has recently gone from knowing nothing about geometry nodes to becoming quite proficient with them, this was the single biggest hurdle I encountered, and the one feature I would love most to see.