Without getting in trouble with the paying clientele, I would just like to say that modeling and trading for mean reversion for interest rate spreads has not for me at least been viable longer term in terms of consistency. I used to trade for mean reversion throughout the '90's and into the early '2000's - but speaking for myself I had to change my ways and develop a more relevant and consistent trade entry methodology several years ago. And it seems to be holding up nicely.
If someone held a gun to my head and insisted that I trade rate spreads for mean reversion, I would look at STIR combinations with more complexity than a simple pair.
The bigger thing I have come to learn about spreading is the importance in the actual construction of the spread combination. That is really a big deal and has a very profound effect on a spread's behavior and modeling characteristics.