Proposal. At least for some period after creation of a new tag, there should be a limit on the number of questions where this tag can be added by the tag creator.
EDIT: As a reaction to animuson's comment and Charlie Brumbaugh's answer I will add that probably it would be reasonable that this feature - if implemented - could be added for a site by request. By default it would be off - in a similar fashion as the 50 questions per month limit is added only for sites which request it. (In this way it would not cause problems for smaller sites which are described in the linked answer.)
I suppose that problems like this occur most likely on sites which are large enough - both to have enough questions and enough users with tag-creation privileges. But on the site that I visit most frequently I have seen something like this happen a few times: A user creates a new tag and starts adding it to new questions. In a few days they retag quite a large number of questions. (Sometimes this might be motivated by badge hunting - a user can get Taxonomist badge. Or this might be done with good intentions.) It might take some time before other users notice it - or even if they notice this, it takes time before somebody decides to do something about the new tag. In some cases, this leads to a discussion on the local meta whether the newly created tag is useful or not might follow - and while this is discussed the tag might still be growing. In the end, the result is quite often that many questions are bumped twice - once by the tag-creator and the second time by the volunteers who are doing the clean-up.1
I think that adding a limit on number of questions with the new tag could prevent such problems or at least reduce them. If nothing else, if the tag is obviously bad and the user in question is just badge hunting, this would be spotted and corrected relatively easily without having to bump too many question. (Originally I have asked about this on Mathematics Meta. Mathematics is the site I visit most frequently. I am not able to judge whether there are similar problems on other sites, but on this site I have witnessed situations where a new tag was added to more than 50 questions within few hours or to more than 100 questions in a few days.)2
I am not sure what would be a suitable value for such limit. (In the linked post I have suggested 5 questions per day.) I do not think this limit would hinder legitimate usage too much. After all, if a tag is really useful, it is going to be added organically; not only by the tag creator but also by other users. And if the user in question is trying to create a tag useful for the site rather than just badge hunting, they will probably not mind that there is limit which represents some kind of "evaluation period" for usefulness of the new tag.
1Well, the tag can still be burninated, thus avoiding the manual retagging. Still, this cannot be done by moderators and requires intervention of Stack Exchange staff. So I suppose this is done only in extreme cases. Other option is that the tag is merged with another tag - this can be done by moderators, but it is only suitable in situation when there exists another tag such that all questions with the new tag fit there.
2Just to be fair to the tag-creator, in the first example I linked above, after some discussion it was decided that the tag might be useful and it stayed. In the second example it seems that the tag might stay, although there is a discussion what would be a suitable way to rename tag (at the moment without reaching some consensus). But still, even if the tag stays, I do not see the need for the urgency to add the new tag to 50 question that quickly. (I think I remember a few other cases where the new tag was added to 50 or more question within few days and later the tag was removed - although I would have to search a bit to find the discussions on meta or in chat about those tags. However, I am not sure whether listing such examples adds too much useful context to the post - all my examples would be related to a single site and experienced users can probably tell whether their favorite sites have this kind of problems or not even without having to look for explicit evidence. So I'll stick with the two examples that I was able to find quickly, since they were relatively recent.)