Not everybody consumes Stack Exchange content the same way.
Some people may want to subscribe to tag uyghurs and be notified by email whenever a new question on the topic is posted. Sure, searching for
uyghurs is:question on Politics SE would accomplish the same. However, if one is interested in, say, 100 topics, who has the time to perform such searches 100 times per day? Roughly speaking, searching does consume time and energy, whereas subscription (say, by email) does not.
If there are only, say, 5 questions on Ruritanians, it is very natural to jump to the conclusion that the tag ruritanians is not needed. However, from the point of view of the tag-subscriber, this conclusion is completely wrong. The more specialized the tag, the more useful it becomes to the (very, very focused) tag-subscriber.
For example, suppose that something "interesting" happens in Ruritania. Mildly interesting. Not as interesting as, say, a coup. Someone who had never been interested in Ruritania may suddenly develop an interest on the topic and post a question with the tag ruritanians. Someone with some expertise on Ruritania, say, a graduate student writing a thesis on the topic, would receive an email notification and perhaps write a high-quality answer while Ruritania is still fresh in his mind. Sure, the grad student in question could search for
ruritania is:question on Politics SE once per year and answer all questions on the topic that have not been satisfactorily answered yet. However, perhaps the ones who posted such questions have not logged in on Stack Exchange in months. Perhaps they will never log in again. The value added by the answer is thus reduced. The experts answering questions do not see their high-quality answers being approved and assume that no one cares about their field and abstain from sharing their expertise. Once the graduate student graduates and gets a "real job", gets married and has children, free time becomes very scarce and his knowledge of Ruritania becomes very stale. With some exaggeration, one could argue that knowledge that is not shared is indistinguishable from knowledge that never existed.