Can’t find any papers regarding this. There can be two different situations:
situation A: A column receives an input, and based on the input (the feature and surrounding columns) the input is (after being processed) routed to a part of the brain specific to that input?
situation B: A column receives an input, and regardless on the input (the feature and surrounding columns) the input is (after being processed) routed to to a part of the brain that sits hierarchically above the first region?
Not quite what you want, but…. “Dendritic Gated Networks” paper, selects a different set of weights based on input at the neuron level (using dendrites), which kind of accomplished the same thing. and way at the end they reference some neuroscience experiments in cerebellum that their method predicts.
It also makes several experimental predictions, one of which we validate with in vivo cerebellar imaging of mice performing a motor task.
Sorry no thalamus magic to see here!
I am also interested in this idea, Brain acting like a router. I thought about making a reference frame game to train agents in, but then had a baby this year 
Different cortical regions tend to respond to different things. Like, somatosensory cortex responds to skin and visual cortex responds to the eyes, but it can be based on more abstract things. For example, there’s a region which responds to intrinsic value, e.g. cookie = food, and there’s another region which responds to learned value, e.g. brands of soda.
Different regions extract different kinds of information from the external world. Maybe that counts as routing, although it might be more like the receiver decides what it wants to receive.