so did we remove 8 images?

Yes and no. We removed the Extra Grid wrapper around the glyphs and collapse the images directly instead. In this case however the test was always flawed. It never tested if the actual image control for this node existed. The change also addresses that issue

Comparing doubles by exact equality can lead to flaky tests; consider using the Assert.AreEqual overload with a tolerance, e.g., Assert.AreEqual(0.5, actualOpacity, 0.01).

This is failing the test

My only comment about moving the xaml code to the code behind is that in the future we should consider to add the controls to NodeView in code behind dinamically based in if they are used or not (so probably this code will need to be re-organized).
e.g. in the image below is shown all the controls inside the NodeView.xaml (we have 21 controls) some of them like customNodeBorder0 or customNodeBorder1 or some Canvas are useless for Out Of The Box nodes (they are used just for Custom Nodes), so I consider we should add them under specific circumstances.

@RobertGlobant20 yes, I think that is what is happening here when we are using the code behind, we check if the node is a custom node for example, and only then add the border. This happens for most controls in the nodeView.

The data context is the viewModel, correct? Why does it need to be null for this function to execute?

This isn't something new, this was here before the change as well, I think we only want to set datacontext once at initialization for nodes

I think is because we are creating the NodeView.xaml (and now InPorts.xaml and OutPorts.xaml) using a DataTemplate (every time a NodeViewModel is found), so the creation of instances is done automatically and the bound with the ViewModel is also done automatically, then the same mechanism applied for NodeView was applied to InPort and OutPorts.