That 'weird' equation does look ok, but it's perhaps a bit of an unusual notation. Can you run
collect_components after that, and let me know if that makes more sense?
The logic behind all this is that Cadabra has the concept of an 'anonymous component tensor', which is what you get when you substitute values for components. Such a tensor is displayed as a 'box' with the indices of the original tensor, and then a list of all components following the large bracket. In this way you can have expressions which are 'half evaluated' (as you noticed in your first picture, where the metric has not yet been substituted).
We are working on making this somewhat less scary to look at, but until that work is done, it usually helps to 'collect all anonymous tensors in a sum', which is what