How difficult would it be to set up custom testnets?
Let's see:
./bitcoin-qt -signet -signetchallenge=007551
./bitcoin-qt -signet -signetchallenge=517551
./bitcoin-qt -signet -signetchallenge=527551
./bitcoin-qt -signet -signetchallenge=537551
...
./bitcoin-qt -signet -signetchallenge=04ffffffff7551
If you create something around 2^32 signets, then I guess some of them will collide with the original one. Also, just by picking "<some push> OP_DROP OP_TRUE" as a signet challenge, you can easily start new signets, and practically never run out of them.
It does not allow you to test things across long periods of time.
If you have a network, which is persistent, then those coins are traded. So, they should quickly disappear, to not get any value on any exchange.
It rather sounds like a complicated way to achieve a temporary regtest network.
In the code, it is a single boolean value, which decides, if difficulty drop after 20 minutes is allowed or not. In case of mainnet, it is simply set to false, but you can change it. By the way: when testnet4 ASIC miners will start reorging CPU-mined blocks, it will have the same effect. And then, if you would be unable to CPU-mine any blocks, then mining on testnet4 or mainnet would be the same, from the perspective of a CPU miner.
However, the regtest starts sounding like the most appropriate way to resolve this problem.
Regtest is too bloated, with the chain speed of six blocks per second. It simply means, that you can produce up to 24 MB/s, and everyone would be forced to handle such traffic.