Non-anonymous users (narrow focus)
In the beginning, making people sysops was pretty casual because all they were expected to do was enforce community bans by blocking people when there was consensus, by blocking vandals or protecting pages that were targets of vandals. These are pretty basic powers and anyone who has been active for a few months or lets say a year understands how to use them. And it is easy to spot abuses of such powers and take them away.
When it comes to recognizing expertise, I think it is more difficult. We are talking about a wide range of knowledges and skills the value of which is much more context-dependent. There are Wikipedians who know I have real expertise in certain areas, either because of formal education or independent research (I think my committment to Wikipedia policy is equally important here). Even so I get into edit conflicts with people who genuinely believe my knowledge is no longer relevant or they know more. This can go to mediation or arbcom. My point is that even thought I believe them wrong, I acknowledge that there is no objective i.e. non-partisan way to handle this. My larger point is that we will often reach the limits of the community's ability to reach a consensus on the expertise of someone.
Look, I know that we do this informally all the time. I do notobject to that. I am just very anxious about trying to formalize it. I really feel safer letting community support work its way in most cases.
Maybe it would help if you explained in as much detail as possible - no need to name names - some of the cases you are thinking of. Obvously you are concerned by some real cases. I think you need to lay those cases out for us, if we are to figure out mechanisms that would have avoided or efficiently resolved the problems that are concerning you.