It is not so academic, if you note, that 02FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFEBAAEDCE6AF48A03BBFD25E8CD0364141 and 03FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFEBAAEDCE6AF48A03BBFD25E8CD0364141 is considered as "zero", when used as "r-value". Which means, that instead of "s=(z+rd)/k", when "r=n", then it is simplified to "s=z/k", and then, the signature no longer depends on "d-value", which means, that it can be valid, regardless of the used public key (it depends on the exact implementation, how it is handled, and if that can cause a fork or not, when one node will consider something as valid, while another node will reject it). Also, it stops being so "academic", if you note, that "Q.x==n" is not the only choice, that can cause some issues. For example, here is another valid key: 027fffffffffffffffffffffffffffffff5d576e7357a4501ddfe92f46681b20a0. And another one: 024ccccccccccccccccccccccccccccccc6b34757867c8fcdeb98be92a3e76ad2d. And then, even if all public keys are restricted, to have their (x,y) coordinates strictly below n-value, then there are still some dependencies inside secp256k1, which can lead you to a "weak" signature, if it was based on a "weak" value like that. Definitely, it is something to be explored on weaker curves than secp256k1, to make sure, that signatures are not too easy to tweak, and that the attacker cannot achieve any advantage, by trying to land on 02FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFEBAAEDCE6AF48A03BBFD25E8CD0364141, 027fffffffffffffffffffffffffffffff5d576e7357a4501ddfe92f46681b20a0, or similar points, for different implementations.
Each and every soft-fork breaks something "by design". It is always about making something invalid tomorrow, which is valid today. Before Taproot, all scripts in the form of "OP_1 <32 bytes>" were valid, as long as these 32 bytes were non-zero. Now, they no longer are, and you need a valid Schnorr signature, or a valid TapScript. The main problem with this specific soft-fork, is that it can invalidate transactions, which were standard for years (for example 2-of-3 bare multisig). Usually, marking transactions as non-standard, and dropping them, is needed, to discourage usage of things, which could be invalidated by a future soft-forks. But if some mining pools will lift more and more rules, then next soft-forks will be more confiscatory, than they were; or they wouldn't happen at all, if their creators wouldn't know, how to make it properly, without breaking some presigned transactions. Also, soft-forks can go very far, and it is known since at least 2016, if not earlier: