BU eliminates the power struggle by unbundling the setting of consensus parameters from the rest of the Code.
Why should this be removed from the code and made user configurable? It is consensus critical since it can create hard forks, so why should it be removed?
It gives choices you say, so does that mean that we should make everything else that is consensus critical and make that user configurable? Should we remove the block reward schedule? Should we change the difficulty retargeting schedule?
Everything controversial, yes, unless we want to vest inordinate power in whichever devs control the dominant implementation. That power concentration opens a major attack vector. I believe this hasn't been noticed before because there was never such a big controversy before. Who knows what the next big debate will be. Whatever it might be, realize that making the choices blockier (less granular) by trying to shoehorn people into one, two, or three possible implementations with bundled consensus parameters doesn't make the situation any better. The market will always choose the least bad option, but if there are only two options, say that offered by Core and that offered by XT, that's quite suboptimal and will lead to less satisfying results no matter what parameter we're talking about.
I agree that having too few options is bad, but at the same time too many options is not optimal either. BU gives too many options for this, and there currently aren't just two options for the block size limit. There are several other BIPs out there with many different alternatives to this.
Part of the problem I have with BU is that I see it as a not-production-ready software. It doesn't having the proper testing (I couldn't find any), and it doesn't consider the all of the consequences. I think it should have been released when deployment options were available and some more options were there instead of just block size limit. There is much more than just the block size limit when it comes to deploying such a fork and not completely screwing up the network.