You've officially changed my mind.
Up until now, I've been harping on the concept of "controlling interest" in which a single entity is large enough to control the direction taken. But I hadn't considered that the new direction might be one that limits the potential for a negative result.
Personally, I think that a sufficiently large instance does represent a major risk. But now I think it's a risk we have to take. If this federation experiment fails, then what is learned can be used in the next experiment.
Now to track down and add a note to all those comments I made...