Yes, but when testnets are no longer worthless, and they get some value, then they start competing with altcoins, rather than testnets. And in that case, if you have a long chain, where coins are traded at 100 satoshis per test coin (at the time of writing in testnet3), then if you can still see, that people are transacting, and moving whole coins without any problems, you can assume, that if mainnet in the future will be worth more, than testnets are worth today, then the network will still be in use.

Also, that fact alone, that even test environments, which were supposed to be worthless, gained some value in the long-term (and are more frequently used in practice, than many competing altcoins), can show you, that things are at least designed properly (and developers have the opposite problem, when trying to make coins worthless, to make a working testnet, instead of trying to increase the value of newly released networks).

Note that even if the hashrate of the whole network will be low, then you can still require a given Proof of Work in your own transaction, by enforcing making a signature, below a given size. So: even if you have 32-bit testnet header, then you can still have 64-bit signature s-value (enforced by the Script), so even if someone will reorganize the chain, then still, that person will be unable to change the chain of signatures, without re-mining them.

Oh, tail supply of course can be made as a soft-fork. Because it doesn't matter, if you increase the total supply, or if you change proportions of owned coins, between network participants. Instead of having a tail supply of 0.01 BTC, people can be simply forced, to send at least 0.01 BTC in fees. And even if moving fees to consensus rules is a bad idea, then it is still better than tail supply (because you can explicitly see, how single satoshis are taken out of all users, and sent to the miners).