Well, one way to start is whenever someone proposes a signaling model ask the following question:
1) What is being signaled?
2) Why is the allegedly signaling behavior a credible signal, either now or in the EEA?
That's a great test of whether a model is actually a signalling model, but not so useful for determining whether any particular signalling model is insightful or true.
It seems to me that when people discover signalling, they see it everywhere and write essays about how no human activity is aimed at its stated purpose.
However, stated purposes and other sorts of useful work get done anyway, and I'm sure there are constraints which mostly keep signalling under enough control that it's mostly not deadly. When I try to think about the subject, I don't get anywhere, possibly because the constraints on signalling are mostly tacit.
Any thoughts or resources on the subject?