in the far future, that means testnet28 will start with barely any block reward, which complicates testing again.
Why do you think, that it complicates testing? Testnets should produce no new coins at all, and allow just pegging things in, testing them, and then destroying UTXOs during peg-out.
How is a new user going to test anything without some coins?
if you could just import some mainnet transaction into testnet
That sounds like a developer-level of testing.
When I started with Bitcoin, I could make cheap on-chain transactions to get a feel of how it works. I must have sent hundreds of transactions between different wallets in my first year (and many of those were zero-fee transactions). It's really too bad I can no longer recommend new Bitcoin users to just try this. I'm not sure how many of them would use testnet for this, and I think most new "Bitcoin users" just keep their coins on an exchange, but if they want to test things, testnet shouldn't be a limitation to them.
Which means, that if testnet28 will contain mainnet block hash at height 12,345,678, then block 12,345,679 can be different in testnet28, and different in mainnet. And then, some mainnet miner can get coinbase reward of 0.09876543 BTC, while completely different testnet28 miner can get 0.09123456 tBTC, by picking a completely different set of transactions, and just getting some fees out of that.
I see where you're going with this. That could work, as long as "testnet28" doesn't find blocks faster than mainnet.
I'm not sure you understood what I said, so if it's divided by 500 million then multiple fuckery by 1 billion.
I was talking about halving the difficulty after 20 minutes without finding a block, so from (say) 500M to 250M instead of the current 1.