But what is the solution if the difficulty reset rule is removed and difficulty remains very high but hashrate drops significantly?
There is none. As other people already said, testnet is defective by design. And I guess developers will keep it that way, because it will encourage people to upgrade, and switch from one buggy testnet to another one, where one problem will be fixed, and another one will be created. Testnets are made to test edge cases, and attacks, which people wouldn't want to test on mainnet, to not accidentally break things.
Some people consider blockstorms as a feature, not a bug. So, if testnet chain will be halted, because of not enough miners, then it may be also considered as a feature, and people can simply switch from one testnet to another. Or: they can test, how coordinated effort works, by trying to unite, and produce the next block. A lot of weak blocks can function for quite long time, and coordinated effort of many parties can lead to making next blocks, moving the chain forward, and lowering the difficulty.
However, that kind of situation is quite unlikely. There are more than enough people, willing to throw their hashrate at testnets, which is why they reached chainworks like 2^70 or 2^75. And when it comes to test networks, people don't want to spend their effort to plan things ahead, but rather they release things on the go, and solve problems, as they appear. Now, there is a problem with CPU-mined blocks, so all of them will be rejected. And if the chain will be stuck, then people will think about that problem, when it will materialize, and not sooner (because many people simply don't care about testnets, and they truly believe, that if coins are worthless, then any fix can be done, including a lot of naughty tricks, which would never be deployed on mainnet).