So i've recently stubled upon an interesting scenario. My workflow looks like the following: http://imgur.com/FERk4CK
The KnimeTest nodes are custom nodes that simply forward table data and have a 10% chance of throwing an exception. I notice that when any of these nodes (besides Node 5) fails, the catch node emits flow variables that contain proper information about what failed. If Node 5 fails, however, the catch node (which has 'always populate variables' enabled) emits flow variables with the specified default values. Is there some limitation to the try/catch functionality when it comes to branching or did i just do something wrong here ?