But this you could do already with the previous setup (use a layer for your reflection objects if you want so)! The problem is that the new setup forces you to create a new layer for each reflection camera. Let‘s say you have 5 objects in your scene that should have real-time reflections via reflection probes. You now have to create 30(!!) new layers and assign them to all your objects simply because you need to have one layer for each of your render targets. That‘s stupid
Damn, why? You would need to create only one layer and place all your objects there.
And cameras too.
You can change renderTarget right before render.
To render a reflection-probe you need to have 6 rendertargets (1 for each face of the cube texture).
How will you render the scene into 6 different rendertargets with the current system without creating 6 layers?
One layer. Many cameras. Many targets. Change target in onPreRender callback.
Not possible either since forwardRenderer calls camera.frameBegin(layer.renderTarget), which affects frustum culling, before it calls layer.onPreRender
Then it’s more a bug in engine, than logical break.