If I were you, Steven, I would also remove non-tradable, illiquid contracts from your list. Stuff like SIMEX Euroyen LIBOR, the Swiss 10y CONF (sadly), LIFFE 10y JGB, etc.
I completely agree with you that liquidity is an important constraint. Unfortunately, I don't really see a good way to go about it. The most liquid markets have so much volume, that at the very least, a logarithmic scale is needed. Total volume is more indicative usually, but there are some markets such as crude oil, where the volume is too spiky to be able to say that the total is indicative of the current. Sessions are another mess. It is easy to say, "Pit Only." But when it comes to volume, it is the volume available during the traders hours that matters. (I considered forming a data company to provide data which take these things into account, but there didn't seem to be enough interest.) A compound issue is that some traders only use correlation/cointegration on lead/lag pairs. For them, the volume on the leading market is of no concern. I agree with you, Martinghoul, but I don't know how to go about it.
I am still working on my "Pair Stability" indicator. Here are some examples: High Stability This high stability/highly correlated pair is only suitable for intraday trading. Modest Stability This pair tracks really well most of the time. Looks suitable for fundamental based, position trading. Low Stability This pair is correlated, but the tracking periods look more like coincidence then economics. You are welcome to try it yourself at my free site http://toolkit.entonesoftware.com/. As this demo site is running on my PC, please don't stack-up a bunch of huge never-ending scans on me.
Sure, I understand the quandary... However, I was only suggesting that you remove things that are obviously not tradable. For example, the Euroyen LIBOR contract on Simex has an OI of exactly zero.