This is a meta meta question on the principles by which we should decide which tags to create. The motivation for the question stems from this meta discussion, in which a majority thought that there shouldn't be an errata tag, whereas a significant minority of $5$ upvoted my answer saying there should be one. Usually that would be the end of it, but after answering yet another errata question I thought again that it's a pity that I can't categorize these although I think it might be useful, and that made me wonder whether there should be symmetry here between votes in favour and against.
This hinges on what the cost of tags is to those who don't find them useful. In the limit where there's no cost, one person who finds a tag useful should be enough to create it and all votes who find it useless should be irrelevant, whereas in the limit where the cost is high, the decision should be by a straight majority vote. Now I'm obviously biased because I'd like this tag to be created, but it seems to me that the cost of creating one more tag isn't very high. We currently have $538$ tags, including quite a number of empty ones, some rather hyper-specific ones like "vapnik-chervonenkis" and slightly informal ones like "simplicial-stuff", and a full quarter of them don't have more than the $4$ questions tagged that I alone could tag right away as relating to errata.
Also, the main use of tags is to be able to find things, either by searching or by following certain tags that you're interested in. Neither of these is in any way affected by there being more tags in total or by individual questions being tagged with more tags -- the more, the merrier. The only scenarios I can think of where it might incur a cost to have more tags is (a) if you're looking at the tags for a question to decide what the question is about and there are so many tags that you can't tell at a glance, and (b) if you're browsing the tags deciding which ones to follow and there are too many to get an overview. (a) seems unlikely to me; (b) might be a problem, but it seems that with $538$ existing tags it's not going to get much worse, and the only solution for that is to browse the more frequently used tags, and of course if the new tag were to get in the way by becoming one of the more frequently used tags then it should have been created anyway.
So my questions are: Where do you see the cost of creating new tags? How high is it? Do you think it's high enough that a tag shouldn't be created when $5$ people see a use for it and $9$ don't?
P.S.: I just noticed http://meta.math.stackexchange.com/questions/2466/new-tags-when, which is on a related topic. (It hadn't shown up in the "questions with similar titles" list). I'm leaving this one open though because it has a different focus. In the case with which that thread mostly deals, splitting or merging general or specific tags, there's an additional cost in that too specific tags might prevent questions from being tagged with the less specific tags (though ideally that should be prevented by a tag hierarchy); that's not the case when a tag like "errata" isn't a subset of an existing tag.
P.P.S.: Another candidate for the "errata" tag just came up: Substitution in ODE.