I'm not convinced this is a problematic bias. What's your prior that the current implementation of LessWrong is near "optimal" in any sense? If it's not, then we necessarily have to Do Something to improve it, don't we? The question shouldn't be if a change is required, but whether any proposed change is for better or worse. And then if it's for the better, is it worth the cost?

Of course, this argument applied even when LW was "doing well". The threat of shutdown is just a convenient excuse for us to participate in the improvement process. It even got a long time lurker like me to post a comment.