Very good idea. There was an article in a website recently using this method (I forgot where I saw it, sorry). It may also be a good idea to always check against a major feed like Bloomberg but that is expensive to do. I have seen wide differences between different feeds, even such that produce negative performance in backtests, The idea of designing feed robust systems is excellent, I do not know if this is possible but I will try if I can collect data for a few of them.