I think it's time for a change.
There will be an official moderator nomination and election 30 days* after the beta ends. For math.SE, this means around end of Novermber.
The nominations for moderator elections will start about 30 days after a site graduates from beta.
Moderator nominations will not be conducted as a meta thread, as they were done with the original sites. We are in the process of creating a custom software module for holding the Moderator nominations. It was the only way to conduct a fair and impartial nomination process, without the quirks of trying to force it into a meta-voting thread.
We just passed 30 days on our first sites. The development is underway so we will start the first round of nominations as soon as the module is ready.
(The voting software is still not completed so it may be later than that.)
There are two reasons why we should be taking the idea of a "change of guard" seriously.
One is that quality of expertise does matter with SX sites, and we might not be making good use of the talent available. By my count, six of the top ten MO users have been active here, and half are 2k+ rep. contributors here, meaning they know this site well, are active, and are fairly likely to accept a moderatorship if it is offered.
The other is that I have reservations about the moderators we do have. Kaestur has not performed badly, but he has been a little bit accident prone, which is really not something we want in a moderator. Katie has generally been solid, but not all that active. Isaac has been good, and KennyTM has been very good. (FWIW, I voted for Katie and KennyTM in the Pro Tempore elections, didn't vote for Isaac, and voted against Kaestur).
Exactly what functions do the moderators perform?
I assume there are automated mechanisms to detect spam and other deletable material that might endanger the operation of the site, and some of this, as well as manually flagged examples that pass the filter, requires human attention.
How much more than that is necessary, given a rating and tagging system and algorithms that can operate on that data? The "crowdsourcing"/wiki model with editing by users also removes much of the need for unique individual powers.
The lack of (1) discussion (2) crossposting (3) threading capabilities -- i.e., the conceptual regression from USENET to the current Stackexchange model -- leads to much of the migration and other work the moderators currently perform. Adding more capabilities seems better than creating a dependence on moderation.
Note: sci.math.research withered in part due to over-moderation, a problem that MO has solved, or rather, solved in the context of a high-end user base. sci.math (unmoderated) would have survived and prospered had it handled the volume growth. SE is vague about how to address these matters.