Well, it would be nice if when someone blocked a bot account, it wouldn't automatically block everyone at that IP address (including the owner).
I don't particularly like the idea of requiring bots to check their own talk pages, at least not on every edit. I think that would waste server resources. It would also significantly raise the bar on bot-writing. If you ask me, we should be *encouraging* more people to write bots, not discouraging them with more red tape. I don't see very many instances of bots gone wild, and surely if it's an emergency, having an administrator block it is quite sufficient control. We can also always run an undo-bot on the account, if necessary to clean up.
I'm also a bit hesitant about allowing third parties to futz with live bots. If there's one bot operator, then there's one person who can be held accountable for all of the actions of the bot. If any of a number of people can come by and frob its code and run the bot, this clear accountability becomes fuzzy.
I'm sure it's useful to post source code for bots, but if a second person wants to use the same source as an existing, useful bot, they should copy it and get a separate account to edit under.
-B.