Just thinking out aloud out the torpedoes, thank you for the exposition. Still there would seem to be lots of possibilities if the DF21 is as accurate as claimed. Sub munitions on airfields would be another good one. I was hoping someone may have more up to date info than I do on it's performance, how it all works and measures to trip it up, especially with targeting.
Developing an accurate weapon is not hard. Developing a long ranged weapon is not hard either. The weapon piece is only a small piece of the big picture.
Successfully executing a detect to engage sequence is one of the hardest things in military operations. Ballistic missiles face a similar problem as trying to hit a high value terrorist running from tent city to tent city. Trying to covertly track a target in an inhospitable environment to the sensor, sending signals without getting caught, and also hoping the target doesn't at any point just bug out and leave or kick up a big cloud of dust and hide. Except the "target" here is much better defended and aware of its surroundings. When the payloads were nukes it didn't matter. With a conventional warhead, it does.
You have to find the target in an environment that is inhospitable to the sensor (anything looking for the CSG...well...then there's things out looking to kill IT). Up to now you've ignored the Blue aircraft flying around, a friendly SAG operating SPY between your scouts and the CSG, maritime patrol aircraft and AWACS from Guam, friendly SSNs, the island of Taiwan itself, etc.
Then that sensor needs to tell a "shooter" where the target is. Over the hundreds, maybe a thousand miles involved, your options are unbelievably limited (options able to operate that far aren't very good), and critical communications nodes are vulnerable to jamming and getting killed. To put it another way, just like operating a CSG in the heart of that land based network is stupid, putting the relatively limited PLAN/PLAAF resources out that far away to search would play into USN hands to snuff out the large sea going units and long ranged aircraft. But you can't execute the DF-21 strikes without doing it, which creates an impasse.
Even if the sensor gets that signal off, over the distances involved, it will take a long time for the weapon to get "close enough" to do its thing. During that time, something has to continue to track the target...without dying in that time.
And in doing so, it needs to update something that is talking to the weapon while it is in transit...because odds are that sensor is not capable of talking directly to the weapon in flight. More signals being sent, more opportunities to be found and die prematurely. While you've got national ELINT assets sniffing just for this type of activity. You keep underestimating the challenges of the comms piece.
You're also hoping that whatever is talking to the weapon (probably a ground relay station) is something cheap that you can proliferate throughout the country...otherwise it's probably already dead from sub cruise missile strikes. But, given the power, stability, and security requirements, it's probably not, and it's probably going to be fixed, and probably going to die too.
If up to this point all went well, the weapon arrives in its "basket" and gets to attempt to execute its purpose in life. Maybe there will be jamming. Or hardkill attempts. Point is, the survival rate of the weapons beyond this point will not be 100% of those that got this far.
But if literally any ONE thing goes wrong in that ridiculously convoluted sequence described leading up to that moment where it becomes directly missile vs target, you just waste very, very expensive missiles splashing into the ocean. A basic principle of war is to find the weakness in that chain and break the sequence. The longer and more complex the chain, the easier it is to find that critical vulnerability.
The long sequence described is why we started using UAVs to go terrorist hunting. With the time from launch to impact, there were too many things that could go wrong to waste a missile. It's optimal if the spotter is also the shooter. And that's when we didn't have to worry about our sensors getting spotted, comms being intercepted, comms nodes being blown up, able to move rapidly with relative impunity (helos and sealift), etc. That kind of operation is hard enough to pull out with all the cards stacked in your favor. Assuming you can pull it off when it's not goes well beyond bold to just plain ridiculous.