This should merge right into d-ronin-next. I'll post the files later today.
I guess whether or not you update the biases in flight depends on your use case. It's a fact that the bias values on these MEMs devices drift, the question is how much, what aggravates it, and is your aircraft flying long enough to be affected by it.
I see your point for freestyle flying, but I think that not updating them might be bad on a platform that flies for 15 or more minutes. But even in freestyle flight, if you get a large bias shift on the gyro that would be noticeable too. So I think you'd still be interested in an accurate bias estimation/correction.
If the biases are updated "correctly" it should be transparent to the end user. Hopefully the accel cutoff will help here.