Obviously, the testnet's block production is unstable, and the old signet is not easy to get sBTC.
What about testnet4? It is more stable than testnet3, and you can get new coins on a CPU. Also, in general, if you move your timestamps 20 minutes into the future, then you can mine blocks with minimal difficulty on testnet3 and testnet4, on a CPU.
Also, when it comes to stability, it seems that some code from testnet4 is also applied into testnet3. And in that case, if you run a new client, then you can get a more stable testnet3. Another thing is that you can get sBTC from official signet, by executing some contracts with OP_CAT:
https://delvingbitcoin.org/t/proof-of-work-based-signet-faucet/937/1Edit: Here you are:
https://mempool.space/testnet4/block/0000000006612e6ecc73e171b07c754d566dad6a8c307298afbdc3723d8982e9As you can see, only 32 leading zero bits are enough, to get a new CPU-mined block in official testnet4. My miner was just lucky, and produced 37 zero bits instead.
By the way, which code are you using for mining? Because when running signet with just OP_TRUE, I could get 50k sBTC, after a few minutes of grinding with Bitcoin Core. Which means, that you are probably running some unoptimized Python code in interactive mode or something, because getting a block with 20 leading zero bits every three minutes is something, which could fit the old Satoshi's prenet from 2008.