MEMS gyros offer low consumption and low cost. How about clustering a bunch together and fuse the measurements in a filter to obtain much better performances. Simon Koldbæk and Luminita-Cristiana Totu did just that in this thesis (the image below is from there) here.
Getting a cluster of 25-deg/hour-drifting-gyros and having a 2.5-deg/hour drift looks great to us.
We're hoping in the future we can integrate the algorithms and do it in real time.